Let’s talk emotions: A case study on affective grammar

English speakers have a choice between the forms ‘let’s talk x’ and ‘let’s talk about x’. We argue that the choice is an example of affective domain intersecting with grammar. To build our case, we explore the contrast between these two forms as it is manifested through syntactic, pragmatic, and semantic constraints. We conclude considering alternative views on idiomaticity and noun incorporation and presenting questions for future research.


Introduction
In both written 1 and spoken discourse, English speakers have an option to produce a form such as (2) instead of (1): (1) Let's talk about books.
This paper evaluates the constraints governing this choice. The existence of such a choice indicates contrast and, wherever there is linguistic contrast, there is a puzzle to be solved. Following Corver (2013;, we view this contrast as a case of language intersecting with affective domain. Corver (2016) maintains that language is an information system and the information it represents must be accessible to the affect system. To that end, otherwise neutral or default linguistic forms may acquire particular, often deviant, forms to convey affective information. This paper addresses such a case in the expression 'let's talk x'. We lay out how the desired emotive impact is achieved and show that while either (1) or (2) can be used to introduce and/or shift a discourse topic, it is only (2) that has the means to unambiguously manifest affective content.
The squib is organized as follows: section 2 consists of a grammatical sketch of the two expressions. First, we survey the pragmatic conditions governing their contrast (2.1). Second, we discuss the syntactic properties of the expressions (2.2). Third, we examine semantic differences between the constructions (2.3). Then we sketch out a theoretical toolbox that would be needed for an analysis of the data (3). We then explain why alternative approaches are not viable (4). An overview of further questions concludes section 5. 1 We focus on spoken discourse. Written data is briefly addressed in Section 5. Glossa general linguistics a journal of Clark, Madeline, et al. 2019. Let's talk emotions: A case study on affective grammar. Glossa: a journal of general linguistics 4(1): 20. 1-12, DOI: https://doi.org/10.5334/gjgl.678

A grammatical sketch of 'talk about x' and 'talk x'
The purpose of this section is twofold. Firstly, a contrastive grammatical sketch of 'let's talk about x' and 'let's talk x' is necessary to show that there is a robust data set attesting to a contrast between the two expressions and that the contrast can be captured through grammatical analysis. Secondly, we need empirical generalizations to build a case for the claim that 'let's talk x' is an example of an affective use of grammar.

Pragmatic differences between 'talk about x' and 'talk x'
This section illustrates the pragmatic conditions for using 'let's talk x' versus 'let's talk about x'. Hashtags are for situationally strange options; question marks indicate borderline options.

Neutral versus affective
In conversation, 'let's talk about x' can be used in neutral or emotionally charged contexts. In contrast, 'let's talk x' is unavailable in a neutral exchange. In (3), 'let's talk x' jars with the calm attitude of the interlocutors. In (4), either expression can be used, although (4a) might be uttered with rage and/or contempt. (3) Context: A married couple is discussing the details of their amicable divorce. a. The choice of expression could be influenced by power dynamics. Switching the roles of patient and doctor from (6), it is unlikely that a young woman visiting a doctor in grave need would start the conversation as in (7a). She would be more likely to use (7b).
(7) Context: A young woman visits a doctor to inquire about an unwanted pregnancy. a. Patient: #Let's talk abortion, doc. b. Patient: Let's talk about abortion, Dr. Smith.

Givenness
'Let's talk about x' allows a speaker to bring a new issue to the conversation or continue with something already figuring prominently in the discourse. However, 'let's talk x' requires an issueto be previously understood or given to some degree-something that figures prominently in public discourse or is available to interlocutors due to their history. This is shown in (8)  We have seen that 'talk about x' works well across a range of pragmatic situations, while 'talk x' is subject to a number of restrictions and dynamic cues contextually or between interlocutors. These observations are summarized in Table 1.

Syntactic properties
We have identified restrictions by context and speaker dynamics. The syntactic properties will reveal the differences in the form of the two constructions. In what follows, we address each of these differences in detail.

Prepositions
The first difference between these two constructions is the lack of a preposition in 'let's talk x'. 2 As an intransitive verb, talk must normally take a preposition for an object to be available. Thus, the lack of a preposition marks the construction as a deviation from the norm. 3

Determiners
The second prominent structural difference between the two constructions is the restriction on determiner use. While the prepositional phrase allows for a noun to be preceded by a determiner or not, the preposition-less version bans the use of any determiner. This is shown in (11a) and (11b).
(11) a. Let's talk about food/the food.
b. Let's talk food/*the food.
This shows that the noun phrase in 'let's talk x' deviates grammatically from the default.

Nominal modification
While the ban on determiners is absolute for 'let's talk x', the ban on modifiers is not. The preference for bare nouns is very strong. However, modifiers are marginally possible, though subject to restrictions. Prenominal modifiers are tolerated as long as they are single-word. Adding more than one prenominal modifier is dispreferred. This is demonstrated in (12a)-(12f). Our current working hypothesis is that while we have similar surface strings that lack P, the factors that condition P drop diverge sharply. 3 There are other constructions with 'talk' that allow for an object without a preposition, such as 'talk turkey' or 'talk sense'. However, 'talk x' allows for variability in object position such that a wide range of variables can stand in for the object. This does not hold for fixed expressions that involve 'talk'-'talk sense' only allows 'sense' and 'talk through one's hat' would not allow 'talk through one's sleeve.' (See section 4.1 on idiomaticity.) c. Let's talk Dirty Harry.
As seen in (14a) and (14b), postnominal modifiers are even less acceptable than their prenominal counterparts.
b. ??Let's talk food that we often cook.

Number restrictions
Finally, the use of number is subject to constraints. Specifically, singular count nouns are not allowed. The plural has to be used, as shown in (15a) and (15b).
As is highlighted in (16a)-(16d), proper nouns are not restricted with regard to number. In these cases, either singular or plural is tolerated. Even though a singular noun implies an individual entity, it includes a plurality of issues associated with that individual. E.g., if we were to 'talk Trump', we would be addressing other pragmatically relevant issues associated with Trump, such as the role of Twitter in politics and Trump's exchanges with North Korea.
c. Let's talk Kennedy.
In short, 'let's talk x' hosts a syntactically reduced noun phrase that includes nominal number and may tolerate prenominal modification but lacks the determiner layer. None of these constraints applies to 'let's talk about x'. These properties are summarized in Table 2. These syntactic properties show that 'let's talk about x' is structurally unremarkable, behaving as a garden-variety prepositional phrase in English. In contrast, 'let's talk x' is marked and subject to a number of constraints and preferences.

Semantics
So far, we have addressed how 'let's talk x' and 'let's talk about x' are similar and different with respect to context (pragmatics) and form (syntax). In this section we address nuances of meaning (semantics).
These constructions are semantically similar in that they are both framed in what is descriptively called the 1st person inclusive imperative (Huddleston & Pullum 2008: 925). In inclusive imperatives, both speaker and addressee are included in the command, in contrast to ordinary imperatives (e.g., 'Talk!' 'Run!').
As shown in section 2.2, the syntactic difference between the constructions lies in the lack of preposition and constraints on the size of the nominal phrase. Semantically, the lack of preposition creates a closer relation between the verb and the noun, akin to noun incorporation. (See section 4.2 for why noun incorporation is an insufficient explanation for the data.) This section focuses on the semantic properties of the nominal within the two constructions.

Common and proper nouns
Neither construction is restricted to just a proper or a common noun as an object. Both options are available in both cases, as shown in (17a)  In 'let's talk x', only bare nouns are permitted. We already know from the syntax section that only mass or plural nouns are allowed and they cannot be preceded by a determiner. However, both complete uniqueness and complete indefiniteness are ruled out. The noun has an interpretation that is in between the two, familiar yet not entirely uniquely identifiable. This interpretation is best captured as specific in the sense of Ionin (2006), where specificity is defined as a universal semantic feature indicating speaker intent and noteworthiness. (19) and (20) exemplify this condition.
Context: A journalist is interviewing a politician and out of the blue says: (a) #Let's talk food.
(20) Context: A journalist is interviewing a politician known for being a foodie and says: (a) Let's talk food. = any food that is somehow relevant in discourse = some specific food that the speaker has in mind, but about which the addressee may or may not share the same assumptions = some specific aspect of food that is prominent in discourse (e.g., they are entering a restaurant) 'Let's talk x' can be used to broach a topic that is current or noteworthy (following Ionin 2006), but may be controversial, such as guns or prisons. When the 'let's talk x' construction is used in the examples below, it involves not only the lexical meaning of the noun, but also anything that is relevant to with respect to the topic at hand. In this case, it is unlikely that the conversation will narrowly revolve around guns or Trump. It is assumed that other contextually relevant issues will be discussed, as in (21) and (22).
(21) Context: The interlocutors are gun aficionados at a shooting range. One says: (a) Let's talk guns. = some aspect thereof, driven or limited by their knowledge and/or experience (22) Context: A well-informed journalist is talking to a politician about current affairs and the politician's involvement with gun lobbying and gun violence. The journalist begins with: (a) Let's talk guns. = some aspect thereof, driven or limited by their knowledge and/or experience Based on this data, the distinguishing contrast between the constructions is the necessary specificity of the bare noun in 'let's talk x'. This is expressed in Table 3.

'Let's talk x' as an emotive topic marker
In this section, we analyze the peculiarities of the form and usage of 'let's talk x' by proposing that 'let's talk x' is an emotive topic shifter. Section 3.1 lays out the tools. Section 3.2 elaborates on our proposal within limits of a squib.

Theoretical assumptions
In this section, we sketch our theoretical assumptions regarding topichood and grammatical manifestations of emotion.

Topichood
The notion of "topichood" is much debated. The most widely used definition is that topic is what discourse is about. But what is "aboutness"? Is it the "totality of coherently related events, states, and referents" (Chafe 1994)? Or an "entity in focus at a given point" ( Gundel et al. 1993)? This conceptual confusion is further complicated by typological diversity. Language-specific manifestations of topic marking do not align easily and may not be comparable. For example, overt topic marking in Japanese does not easily compare to English, where there is no single obligatory form that marks topics (Givón 1983). The controversy is beyond the scope of a squib. For a broad overview and discussion of these issues, we refer the reader to Büring (2016) or Riou (2015). For this narrow case study, we define topic or aboutness as contextual noteworthiness that is grammatically marked (Ionin 2006;Bergen & Hoop 2009;Riou 2015). In other words, we assume that topichood must be unambiguously indicated by a grammatical manifestation.

Emotion in language
Emotions infuse language. Laymen and scholars have been fascinated by manifestations of emotion in language since Aristotle. There exists a wealth of literature on emotionally driven choices as far as lexicon (e.g. Wierzbicka 1986;Pavlenko 2007), morphology (Grandi 2015, a great recent compilation), and prosody are concerned (Frick 1985;Wichmann 2000;Edwards et al. 2002). A long tradition in semiotics of language is surveyed in Wilce (2009). Moreover, there is a growing subfield of computational linguistics exploring emotion through internet language use (e.g. Tokuhisa et al. 2008;Recasens et al. 2013;West et al. 2014). Traditions grounded in Saussurean structuralism and formalism did not give emotion much thought, with a few exceptions such as Roman Jakobson (1960). Thus, there is little literature on how emotion interfaces with the system of grammar ( Corver 2013;Majid 2012). This may be due to the richness of emotional encoding in lexical and prosodic domains, much of which remains unexamined. It may also be that few linguists look for evidence of emotional impact in grammar because they perceive grammar as a purely computational module. Along with Corver (2013;, we beg to differ. There is no reason grammar as a system should be immune to emotion. Consider focus and evidentiality, which are essentially manifestations of emphasis and veracity. One conveys a perceived prominence in discourse, while the other reveals the speaker's level of commitment to a proposition. Neither focus nor evidentiality must be grammatical or structural, but they may be. Thus, syntax must be able to accommodate them. There are attested cases of languages in which focus can be fixed structurally (Farkas 1986 on Hungarian, and subsequent work), or evidentiality is structurally manifested via morpho-syntax (Speas 2008). Thus, it is not a stretch to expect structural evidence of affective impact. Because language is laden with emotion, models of grammar cannot escape its allure or ignore its presence. Following Corver (2013;, we assume that affective content seeps into the grammar as deviations from "default" representations. Corver notes that these deviations can be perceived as marked constructions because they are used in "non-core-grammatical" ways, or relax some conditions of the core grammar. Straying from default forms is where affective content takes hold. Based on his study of Dutch, we assume Corver (2016) means these "non-core-grammatical" ways of syntactically signaling affect may be language-specific.

Proposal
Our data and speaker judgements suggest that 'let's talk x' is a manifestation of affective content. As described in section 2.3, 'let's talk (about) x' is a 1st person inclusive imperative. It contains an accusative form of we (contracted to 's) whose reference normally includes both the addressee(s) and the speaker. At least two interlocutors are signaled in the form of the expression. Further, the presence of 'talk' in the imperative mood complements this grammatical frame in highlighting specific topics-subjects that have noteworthy aboutness (as in Ionin 2006). In that sense, 'let's talk about x' and 'let's talk x' are similar.
The difference is that 'let's talk x' is obligatorily emotive while 'let's talk about x' can be used emotively but does not have to be. Emotive content consistent with 'let's talk x' could be conveyed by 'let's talk about x' through prosodic cues but emotion is part of the structure of 'let's talk x'. Extending Corver's approach, we propose that the grammatical deviation may be a constellation of factors: impoverished imperative structure, lack of preposition, bareness of noun, and restrictions on modifiers.

Alternative approaches
This section addresses two alternative interpretations of 'let's talk x', both of which would view this construction as a fixed expression, not a part of productive grammar. Section 4.1 discusses idiomaticity while Section 4.2 discusses noun incorporation.

Idiomaticity
An idiom-based view does not hold. Idioms are phrases learned by rote in their entirety and are usually not compositional. One could argue that there is a fixed inclusive imperative format characteristic of 'let's talk x'. However, this format is grammatically driven rather than lexically idiosyncratic, compared to idioms such as 'cut corners' or 'cry over spilt milk'.
Crucially, 'let's talk x' can take a wealth of nouns in the relevant position. In contrast, 'cut corners' and 'cry over spilt milk' could not retain their idiosyncratic meaning if they morphed into 'cut regions' and 'cry over spilt wine'. Moreover, as one can see in the literature-perhaps due to McGinnis (2002)-many authors view the study of idioms as part of the study of grammar (Anagnostopoulou & Samioti 2013;Harley & Schildmier Stone 2013).

Noun incorporation
Another view is to treat 'let's talk x' as noun incorporation into a verb, akin to 'truck-driving individual', where 'truck' is incorporated into 'drive'. This is appealing at first glance since the verb stays fixed and incorporates an array of nouns. However, it falls apart under closer inspection. Our counterargument relies on a detailed overview of noun incorporation by Massam (2009) and insights by Gehrke and Lekakou (2013). We will not address the morphosyntactic intricacies of noun incorporation here because, due to the diversity of data, there does not exist one agreed-upon view of noun incorporation. However, there appears to be agreement on the semantics of incorporated nouns, which are non-referential, modificational, and sometimes classificatory. In 'truck-driving individual,' 'truck' is not referring to a particular truck: it modifies the verb 'drive' to convey a type of driving and classifies it. In 'let's talk x', the noun does not modify or classify the verb and is referential. Noun specificity is required. Lastly, noun incorporation is a well-known, commonly attested phenomenon. It is not deviant in the sense of Corver (2013;.

Conclusions and further questions
This squib was inspired when a student asked about the difference between 'let's talk about x' and 'let's talk x'. Assuming that contrast in structure was significant for the distribution and interpretation of these structures, we assembled pragmatic environments, modifiers, and determiners to check for noteworthy differences. We found robust differences, which are intuitively available to native English speakers.
Relying on the grammatical frame of the constructions-inclusive imperative complemented with 'talk'-we posited that both constructions are used to introduce topics (what current discourse is about). However, 'let's talk x' is marked as emotionally charged and dedicated to a narrower set of specific or 'in the air' topics (specificity as in Ionin 2006). Finally, adapting Corver's (2013; views on how affective domain intersects with grammar, we proposed that emotional charge may arise from grammatical deviations from a standard form.
While we captured the intricacies of deviation for affective impact in this construction, we unearthed more questions. Some questions are specific to this construction, while others spill into larger concerns.
One unanswered question with respect to 'let's talk x' is why and how we get a sense that expertise in the relevant subject matter is expected. It would be pragmatically strange to talk about something that neither interlocutor has experience with.
The second question concerns structure: how to account for (i) drop of P; (ii) and yet allow for the presence of noun argument with an otherwise intransitive verb? One possibility would be to argue for (silent) incorporation of P into V (following Hoekstra 1988; 5 see also Hall 2018).
Another set of questions pertains to written data. A Google search reveals examples of 'talk x' in newspaper headlines ("Pelosi to talk policies") and advertising ("Let's talk drop-in consultation"). It is even used in shop names ("Let's talk travel", spotted in Brighton, England, spring 2017). We are unsure if the impact and affective content is the same, and how, if at all, addressing a presumably unknown reader alters the nuances outlined in this study.
A larger question is how one formally captures the range, limits and types of deviation that signal affect. Will any deviation do or are only particular deviations appropriate? Do thesedeviations have limits to what they can do or which affective flavors they can manifest? Do they, as in 'let's talk x', involve a constellation of factors in form, meaning and use? Can one predict types of deviation and how they map onto grammatical systems? Are there universal patterns or constraints? How are affective manifestations different from lexically encoded emotion ormechanisms in the iconic domain such as onomatopoeia and interjections? We hope to study the affective domain further and urge the reader to consider the questions raised here.