Introduction to Garfinkel’s ‘Notes on Language Games’: Language events as cultural events in ‘systems of interaction’

Why a third version of the Trust argument?

Each version of the Trust argument addresses a different aspect of the same sociological problem: how to convincingly argue that an adequate theory of social order cannot be achieved as long as sociology treats social objects/events as if they were ‘natural’ when they are created through an orientation to constitutive conditions. In a move that parallels Wittgenstein, Garfinkel pointed out in 1946 that the problems that result from that mistake, like the difficulty explaining social change, taking language for granted, and the emphasis on motivation, have been mistaken for problems intrinsic to social order, when they are merely artifacts of the initial mistake.

The earliest versions of the Trust argument in the 1940s maintained that adding constitutive conditions (with particular regard to language) was required in order to make Parsons’ approach work, a point Garfinkel made again in 1962 in Parsons’ Primer (Garfinkel, [1962] 2019) in a slightly different way because Parsons’ approach had changed by 1960. The objective was to eliminate contradictions in Parsons’ theory of ‘interactional systems’, which Garfinkel maintained was flawed by assumptions about language that prevented Parsons from seeing the constitutive character of interaction. In addressing this issue, Garfinkel had turned to Schutz for a sketch of what the taken-for-granted practices in question might look like.

By 1960, however, Parsons had seen the problem and was attempting to incorporate a constitutive approach to interaction/culture into his pattern variables (although not having focused on how language actually works, Parsons was still trying to limit and control the variables, so his approach still fell short of what Garfinkel required). That Garfinkel turned to Schutz for descriptions of ‘taken-for-granted’ practices in an attempt to illustrate what an ‘interaction-in-its-course’ in Parsons’ interactional system would look like, led to the argument being misinterpreted as phenomenological. But Garfinkel was using Schutz’s descriptions of taken-for-granted practices to address a shortcoming in Parsons’ conception of interactional systems – a constitutive – not a phenomenological position.

This third Wittgensteinian version of the argument lays the foundation for Garfinkel’s approach to language, while also addressing developments in cultural/linguistic anthropology that were threatening to undermine Garfinkel’s interactive approach to ‘structure’. A group of influential cultural anthropologists who were studying culture in terms of linguistic classification systems had adopted the name ‘ethnoscience’ for their approach. These scholars, who included Ward Goodenough, Volney Stefflre, Charles Frake, Duane Metzger, and Cornelius Osgood, had a presence at UCLA and UC Irvine, along with Yale and Chicago. ³ From Garfinkel’s perspective, they were proposing to take the study of culture backwards by popularizing a cognitive/semantic conception of language and its relationship to symbolic classification systems that undercut his own approach to ethno-methods.

Similarities between Garfinkel and Wittgenstein afforded a platform from which to defend against this development. In Garfinkel’s view, the relationship between culture and language was made evident in the use of language, in how language events meet constitutive criteria, and could not be found in static and abstract symbolic classification systems (a neglected argument Durkheim had also famously made, see Rawls, 1996, [2004] 2008). This had also been a basis of difference between Garfinkel and Clifford Geertz, another of Parsons’ PhD students in the 1950s, and Garfinkel had been working with Parsons since 1958 to bring him closer to a fully interactional view of culture. While their collaboration on the revision of the pattern variable argument (Parsons, 1960) did make that argument more interactional, the regressive stance being taken by cultural/linguistic anthropology threatened to undermine that progress, especially as they had appropriated the same name: ‘ethnoscience’, that Garfinkel had begun using in 1953 for his own work: It had to be addressed.

In ‘Notes on Language Games’, Garfinkel launches a challenge to these cultural anthropologists, along with the game theorists, Von Neumann and Morgenstern, with whom he had debated at Princeton in 1951–53. In launching this challenge, Garfinkel adopted a Wittgensteinian framework that allowed him to treat games, language, culture and interaction as involving the same essential constitutive commitments and processes. His objective was to challenge the semantic/conceptual perspective on language that these cultural anthropologists were assuming, while at the same time challenging the conception of rules and rational game play, and their separation from ‘serious’ everyday life that game theorists assumed.

What Garfinkel added to Wittgenstein was a conception of constitutive conditions as determinative of any field of social action – whether play or serious – along with an insistence on studies of actual concrete language use. He maintains that the way constitutive conditions work in games is how they work in serious life so that understandings derived from studying constitutive practices in games can be applied to social interaction/culture without limitation.

John Austin (1962) had attempted a constitutive argument about ethics in writing about Aristotle in 1939, but it remained relatively obscure, and even when picked up by Alasdair MacIntyre ([1981] 2007: 223–4) in After Virtue, was not treated as a constitutive practice argument. John Searle (1964) came closer to the kind of argument Garfinkel was making in proposing constitutive criteria for Speech Acts. But the argument retained elements of conceptual abstraction and continued trying to limit and control variation like Parsons had with the pattern variables. John Rawls (1955) came closest to what Garfinkel meant in ‘Two Concepts of Rules’ (see also A. Rawls, 2009). The idea that constitutive criteria can be used to produce certainty from contingency in language and social practice, however, never really got across. Durkheim faced a similar problem with his constitutive practice argument (Rawls, 2008; 2012; 2019).

Garfinkel’s argument in ‘Notes on Language Games’ ties up all of these traditions. Near the end of the manuscript he outlines the theoretical implications that follow if, as his initial research had indicated, language and cultural events all have basic rules that operate in the context of his three constitutive Trust conditions. This finding, he says ‘leaves open the immensely important possibility that the constitutive accent is an integral feature of all events irrespective of whether they are events in the domain of a game, scientific theorizing, theatre, play, dreaming, or whatever’. In other words, Garfinkel is proposing that constitutive conditions are the missing foundation of social theory.

It is unfortunate that the manuscript was only partly finished (with extensive handwritten notes that are difficult to read). Demonstrations and theorems that it promises are also missing. Some unfinished notes are included in two Appendices to the paper, as published in this issue. These include references to Wittgenstein (Appendix I) that were in the notes, but not in Garfinkel’s text, making it possible to locate with some precision the sections of the Philosophical Investigations and Blue and Brown Books that Garfinkel worked from.

Language events and cultural events are the same

In saying that language events and cultural events are the same, Garfinkel means that both must be accomplished over a ‘course-of-interaction’ that is oriented toward constitutive expectations such that they can be recognized by other participants as having a particular meaning in a particular sequence of interaction. Garfinkel does not privilege language events by semantic or dictionary meaning. Neither does he privilege cultural events by beliefs and folkways. The so-called durable objects of culture and language all need to meet constitutive requirements as actual concrete interactions-in-their-course.

Garfinkel did not deny that words have conceptual or semantic meaning. What he said was that they have too much meaning. The problem on any actual occasion of use is to figure out how participants make any particular sense of a word or saying. In 1948, Garfinkel proposed that a sequential process must be involved in which the participants orient constitutive expectations regarding the significance of features of sequential placement (Garfinkel, [1948], 2006). That process of back-and-forth exchange of actions and sayings (which Garfinkel later called ‘reflexivity’) is similar to what Parsons would in 1951 call ‘double-contingency’. But Garfinkel’s formulation already relied on constitutive expectations in 1946, while Parsons only came to that conception around 1960.

The big point Garfinkel makes is that studies of language are sociological in the same way that studies of action, interaction and culture are sociological. He restates the problem of social order as the problem of linguistic order – a point that has been difficult to get across to sociologists. Unfortunately, this focus on how culture and language are assembled was early on misinterpreted as a focus on a contingent and individual level of analysis. It is not. Garfinkel was filling in the empirical detail of the social system of constitutive interactional practices that Durkheim and Parsons had argued for. This mistaken identification of ethnomethodology with individualism was due in part to an identification of the Trust argument almost exclusively with Schutz and phenomenology that neutralized its sociological significance. That the argument conflicted with what Wittgenstein called ‘the craving for generality’ also encouraged misinterpretation.

From the first, the key to Garfinkel’s approach to interaction as an organized structure of social commitments has been the three constitutive Trust conditions.

Constitutive practices change the question of correspondence

In papers from the 1940s and in his 1952 dissertation, Garfinkel made an argument that is similar to Wittgenstein’s criticism of the Augustinian picture theory of language. Garfinkel’s point was that assuming that language ‘just’ works through a correspondence between word and meaning or object is a mistake that has stopped sociologists from looking at how language works. Garfinkel called this mistake one of four pre-theoretical decisions that can set sociology off on the wrong path.

In this manuscript, Garfinkel appears to reverse his argument, stating that there can be a correspondence between intended and achieved meaning in the context of legitimate constitutive criteria. His position has not changed, however. The difference turns on the distinction between a social context defined by mutual commitment to constitutive practices, in which meaning is achieved by orienting an implicit shared ‘social contract’, and a context comprised of natural individuals and natural objects with no constitutive commitment.

Garfinkel’s point in ‘Notes on Language Games’ that correspondence can be observed rests on his argument that the achievement of meaning in a constitutively bounded social context is evident in the way a saying is treated by the other participants. Whether it has been successful or not is an empirical matter. In this sense, ‘intention’ has become a feature of the constitutive criteria an action must meet that the other participants can observe, and is not an idea or concept in the head.

In Garfinkel’s terms: ‘A sign “correctly” corresponds to a referent in terms of the assumed constitutive order that itself defines “correct correspondence”.’ Sanctions (and self-corrections) can be expected to follow incorrectness as empirically observable matters.

‘Structure’ is rule-governed action within a social contract

Garfinkel uses the terms ‘structure’ and ‘structures’ in this manuscript to refer to observable features of rule-governed actions, including sanctions, as noted above. This is not how the term is typically used. Structure, for Garfinkel, refers to the feature that actions which orient the same constitutive criteria can have a routine repetitive character – not because they are generalizations that abstract from the detail of particulars – but because the persons who create those actions do so with an orientation toward shared constitutive criteria. That is, their routine character is the consequence of being made by participants to be recognizably the same. In order to convey shared meaning, participants need to have their actions seen by others as particular actions according to witnessable (empirical) constitutive criteria. Garfinkel refers to this as the ‘oriented’ quality of action (a term of Garfinkel’s that appears in Parsons’ revised pattern variable argument). The recognizable empirical features of actions that orient constitutive expectancies he calls structures. Structures, in such usage, are not like class relations, formal institutions, or ‘linguistic structures’ like classifications.

The structures Garfinkel refers to are the observable constitutive features of language and action produced when participants orient the constitutive expectancies of language, culture, and interaction in trying to make sense for one another. The uniformities of language that result Garfinkel calls structures in the same way that interactional and game events are structures when they are recognizable according to mutually oriented constitutive criteria.

Using games to illustrate points about language and interaction/culture

Garfinkel promises to demonstrate a formal correspondence between literal descriptions of the constitutive properties of language and games in six theorems. Unfortunately, we only seem to have three of the theorems in the unfinished manuscript. Literal description he formulates as a step-by-step description of how a linguistic or game event is achieved. Because the process is essentially the same for language and games, he suggests it will be easier to begin with games.

Garfinkel states two theorems about the correspondence between language and games. Theorem one says that: ‘Actual or observed uniformities of linguistic events – i.e. actual uniformities – whether they be uniformities of “speech events” or “meanings” – consist of institutionalized possibilities of usage which comprise the language as a legitimate order.’ He explains that: ‘In Wittgenstein’s phrase, the language is a paradigm of intended possible usage, i.e. language is a legitimate order.’ Theorem two says that: ‘The stable features of linguistic structures are guaranteed by motivated compliance to the language as a legitimate order, i.e. as a paradigm of intended usage (cf. Wittgenstein).’ Both theorems cite Wittgenstein.

With these two theorems Garfinkel covers both the institutionalized possibilities of use, which Wittgenstein called a ‘paradigm of intended possible usage’, and the production of compliance and legitimacy through sanctions (an idea derived from Parsons and Weber). The reproducibility of these structures is guaranteed by motivated compliance to a set of possible actions that actors orient, which are institutionalized through implicit social contract, and about which they expect sanctions to occur when criteria are not met and meaning is not achieved. Both conditions are necessary, he argues, if intended and achieved events are to stand in any evident correspondence.

In addition to restating aspects of Garfinkel’s own Trust conditions, these theorems are reminiscent of his formulation of the relative contributions of Parsons, Durkheim, and Weber, to the sociological conception of legitimate order in Parsons’ Primer (Garfinkel, [1962] 2019).

The subjective/objective problem in a context of constitutive practice

Garfinkel points out that when everything is created within an implicit social contract, by participants who are orienting the constitutive requirements of that contract, all meaningful actions and objects are ‘objective’ in the sense that they are publicly available to all competent and committed participants in the constitutive practice in the same way. Even what we refer to as ‘subjective’ meaning is attached to events using this process.

Things that are simply personal – or that fail to meet constitutive criteria – have no public meaning at all. Looked at this way, most considerations of the subjective/objective problem are backwards. The trouble, Garfinkel says, begins when people start ‘dividing things up’, by which he means doing things like treating linguistic classifications as ‘durable’ social objects and then taking them out of context and classifying them for analysis. If objectivity is only achievable within a constitutive context, then dividing things up conceptually, which takes them out of context and out of interaction, is a problem: It starts the mischief.

Garfinkel prefers a ‘scenic reference’

Keeping objects and social practices in context, and not abstracting them from their context, involves what Garfinkel refers to as ‘scenic reference’. There have been arguments over the years about the difference between ethnography and ethnomethodology, phenomenology and ethnomethodology, and between what Garfinkel called ‘formal analysis’ and ethnomethodology. In each case, the difference has to do with how the empirical material is approached and not with its content per se: But for Garfinkel, the approach can alter content in significant ways. Whether it is a transcript of a conversation, the experience of a ‘way-finding journey’, or the observation of others at a workplace, Garfinkel insists on preserving what he calls the ‘scenic’ features of the event. What we need to focus on is what the people in that place who made these meanings had available to them at the scene to orient as constitutive resources with which to make meaning. As a ‘way of working’, Garfinkel and Sacks (1970) made a commitment to confining themselves to the constitutive resources participants had available to them.

Garfinkel maintains that emphasizing embodied and situated action at the scene where people are enacting the constitutive features of a social contract avoids the problem of conferring physical locality – or locality within an individual – to constitutive features (like rules and meanings) that are not locatable in these ways, and references Wittgenstein for the point. If the focus of observation is on actual concrete constitutive practices, then it is essential to stay focused on the scene, or situation, where those constitutive features are located; and they are located only with reference to the criteria that make them real and nowhere else.

Constitutive versus basic rules: the Trust argument

Because, in Garfinkel’s view, everyday life, culture, and language are like games, he can begin by demonstrating how a constitutive social commitment/contract works in games. According to Garfinkel, a combination of basic (constitutive) rules and preferred rules defines all games, cultural events and language events/games. ⁴ The Trust argument specifies required orientations toward these rules. As a social form with basic and preferred rules, games are, in Garfinkel’s view, a simple society, unlike the way game theorists describe them. Furthermore, they are useful for research because, he says, they are ‘familiar’ and new ones ‘can be invented’ to fit the purposes of inquiry. Garfinkel, near the end of the manuscript, identifies games with serious life:

Should it turn out that the boundaries of the set are specifically vague; that the set is always a partially ordered one; that every game has what Durkheim called ‘unstated terms of the contract’; and that timing is a parameter of the meaning of a behavior, then we have important grounds for optimism, for those are precisely the properties of those situations of events that sociologists have referred to as the actor’s ‘definitions of the situations’ of ‘serious life’, and that inquiries have documented to the point where these properties may be safely assumed.

Preferred rules, in contrast to basic rules, are selected by participants from a set of possibilities. How they do this involves options, but not just any options. Options can be preferred or dispreferred (or may become rank-ordered as to preference as in turn-taking), and some options, even though not prohibited, may call the user’s competence into question.

All game events are members of the set of either basic or preferred rules. But they are typically distributed between the two sets. The case with only preferred rules and no basic rules does not exist, Garfinkel says, and: ‘The case in which all possibilities are constitutive possibilities, such that the sets of preferred and conventional possibilities are empty sets, defines a ceremonialized game.’ This is a game with constitutive requirements that leave no options. Garfinkel argues that there is no need for a conception of the ‘sacred’ (which typically focuses on ceremonialized games) to explain the moral accent on sanctions in games (Rawls, Jeffrey and Mann, 2016). Not only is there no need for the sacred to explain the moral accent, but the conception of the sacred in this regard suggests a set of sacred rules that are ‘more fundamental’ than the basic rules, and Garfinkel maintains that there is no such set. It is violations of the basic rules, he says, that lead to moral indignation:

The alternative reasoning is this…the critical phenomenon is not the ‘intensity of affect’ with which the ‘rule’ is ‘invested’, or the moral status of the rule, but the perceived normality of the scene of events, as this normality is a function of the presuppositions that define the possible events.

Senselessness versus ambiguity

An action that breaches the basic rules is ‘senseless’, according to Garfinkel. It has no necessity. It cannot be recognized. ‘It acquires’, Garfinkel says, ‘the perceived properties of unpredictability.’ Because it is an orientation toward constitutive criteria that makes anything seem routine, standard and predictable, when actions breach basic rules, they appear to be unpredictable. Not understanding that this is an effect of breaching basic rules has led to the mistake of thinking that social action has the general characteristic of unpredictability. When people are playing without rules – or not playing by rules – actions will seem unpredictable. But game play does not in general have this character and neither does everyday life.

This does not mean that the actions of others are ever ‘predictable’: that is another mistake one can make while doing sociology. It means that if actions recognizably orient constitutive rules, they will not have the ‘perceived character of unpredictability’, a very different thing.

In explaining this, Garfinkel proposes that:

A distinction between ambiguity and senselessness may help the discussion. By saying that the field of game events becomes ambiguous, I mean that the player’s distribution of bets as to ‘what happened’ over the set of alternative possibilities becomes more equiprobable.

Ambiguity involves not being able ‘to decide which among a set of alternatives a person meant in a move, or an utterance’. Senselessness involves having no choices. According to Garfinkel, ‘In a senseless field, the person, although he/she hears an utterance that has been delivered in clear and correct English, does not recognize it as an English sentence. What holds for an utterance holds for any behavior…’

An action is a ‘course of transformations’ of expressions

Garfinkel maintains that ‘Preferred rules are procedures for transforming environments of game events in accordance with considerations of’ how to accomplish future states of the game. That actions are procedures for transforming the environments within which they occur is only the case, given a shared commitment to constitutive practice. Garfinkel makes a similar point about meaning transformation in Parsons’ Primer, arguing that Parsons’ revised pattern variables are procedures for transforming the meaning of expressions.

All game events, Garfinkel says, are sign-functions and all sign-functions are expressions. Basic rules define the domain of action. Actions are the operators. Sign-functions are, Garfinkel says, ‘methods whereby environments of linguistic events are transformed’.

Conclusion: Garfinkel in 1960

The period during which Garfinkel wrote ‘Notes on Language Games’ was positively saturated with intellectual projects and opportunities. Having spent the two years immediately after completing his dissertation chasing jobs from Princeton, to Carnegie Mellon (which didn’t materialize), to the Ohio Research Board, to the Jury Project at Chicago, and then finally securing a tenure track position at UCLA (only to be told they were letting him go because of complaints about his teaching), Garfinkel was finally somewhat secure. ⁵ By 1957, he had not only succeeded with his teaching, but had secured a large Army Air Force grant that would support his research over the next five years. This supported research at the UCLA Clinic with Bittner, interviews with 15 transsexuals in 1958–59 (including ‘Agnes’), two seminars on Parsons in Spring 1959 and 1963, and Garfinkel’s work with Sacks at the Suicide Prevention Center. The funding also gave him the freedom to spend extended periods at Harvard with Parsons, and in New York with Schutz.

1960 was an important year in the middle of this period of intense productivity. Garfinkel had reconnected with Parsons in 1958 and was working on Parsons’ Primer. He had met Sacks at Harvard in 1959 and the two began talking in the Spring of 1960. The transsexual research was underway, and soon Goffman would write asking Garfinkel to publish a book with him on stigma and ‘passing’. The Trust argument which had undergone so many revisions (being submitted and rejected more than a dozen times in 1957–58 alone) had finally been accepted for publication by O.J. Harvey. The research on ethno-methods that Garfinkel would soon become famous for was proceeding apace. The clinic study was underway. The Jury study was complete.

Scholars not familiar with the density of Garfinkel’s scholarship during these years have argued that there was a change in his position involving a switch in allegiance from Parsons to Schutz (or from Schutz to Parsons depending on the misunderstanding involved), and that his position became more ‘radical’ over time. In light of these misconceptions, it matters a great deal that the work in question was all being done at the same time. While he was preparing the Trust argument based on Schutz (from 1956–63) and the Trust argument based on Parsons (from 1959–63), Garfinkel was also preparing this version of the Trust argument based on Wittgenstein and the idea of language games in March 1960.

Having spent time with Garfinkel in his study later in his life and seen him at work, I can imagine that he applied himself to all of these projects at the same time. In the 1980s and 1990s, he had notebooks – eventually 14 of them – spread out on long tables and would roll his chair from one to the other, reading and making notes. Not having been there in 1960, I can’t be sure that his procedure was the same. But the materials are preserved in the Garfinkel Archive in notebooks – blue cloth binders in some cases – that could have stood open on his tables just as the 14 notebooks did later, and it is clear from the dates that he was working on them at the same time. Whether or not he rolled his chair from notebook to notebook in the 1960s, the arguments of the many projects he was working on overlapped in significant ways just as they would later.

In the Parsons’ Primer manuscript, Garfinkel explains that it was Parsons who first put sociology on a social contract footing in which the rules for assembling interactional practices provide the framework for whatever would be recognizably done and what it would mean. In that manuscript Garfinkel offered his Trust conditions as the missing implicit commitment that was necessary to ground Parsons’ system of interaction/culture and offered his own early research as the missing empirical demonstration of Parsons’ position. At the same time, he described Schutz as offering insight into how the assembly practices of culture/interaction might be described in process, so to speak, as assembly practices in-their-course. Rather than representing a change in Garfinkel’s thinking, the references to Parsons and Schutz go together. They are the same argument, with Schutz supplying the initial detail for Parsons’ framework.

Wittgenstein supplied a third approach to the same issues. While Schutz had helped him begin thinking about what actions in-their-course might look like, Garfinkel wanted to focus on constitutive practices in a more literal way. In his paper on language games, Garfinkel describes Wittgenstein’s method as ‘literal’, by which he means that his step-by-step descriptions of how language games work are meant to be adequate to explain and reproduce them in use.

Two things had come together by 1960 at UCLA, one positive and one negative. First, the theory/method Garfinkel had been crafting for many years was finally bearing fruit. He was able to describe with increasing precision just how social objects were being assembled in interaction. He was becoming so confident about this that he began offering his own research as a defense of Parsons’ theory in 1963: both in letters to Parsons and in the Primer manuscript.

Second, Garfinkel found himself surrounded by linguistic anthropologists at UCLA who were using his own terminology – ethnoscience – to take a conflicting cognitivist, semantic, and grammatical approach that ignored how people used words on actual occasions of use.

In Garfinkel’s view, all social objects – including words and meanings – are created by participants using shared assembly practices that orient legitimate and sanctioned systems of interaction (what Goffman would later call Interaction Orders). Neither linguistic anthropology nor the new linguistics (by Chomsky) ⁶ paid any attention to this level of order. Wittgenstein, by contrast, considered how words are actually used (even though he only did so reflectively) and argued that this was the heart of the matter. It is not at all surprising that Garfinkel turned to Wittgenstein at this point. What is a bit surprising is that the manuscript remained unknown.