【按】如何精读与细剖一篇理论文本？如何不像只是summary地读？如何触摸文本的“肌理”？如何既constructive又critical？

——这可能在多数做社会学理论研究的常见困扰。

——为此，赖得吕炳强老师赐稿，下面贴出，他是如何解剖《Sociological理论大缸》【下部】推送过的一篇AJS发表的关于民族志中机制解释的论文。

——第61期：如何【精读细剖】一篇AJS理论文章？来次示范吧！（上部）

——第56期：民族志也能【做出】机制式因果解释？实用主义回答.（下面评论的那篇AJS）

——第27期：开学季！课程大纲《社会学实践的哲学规章》（这是吕炳强老师在16fall于CUHK开设的一门课程。）

______________________

Commentson A Pragmatist Approach to Causality in Ethnography, by Iddo Tavory and StefanTimmermans（AJS,2013,119 (3): 682–714）

 以下P.+num.是 这篇AJS论文的页码。Lui即吕炳强老师的剖析文字。

P. 688:

FROM SEMIOTICS TO VARIATION

Peirce’s semiotics provides us with an analytical tool settotrace sequences of action at a foundational level. However, as philosophers havealready observed, the focus on mechanisms does not solve the metaphysical problemsof causality, and needs recourse toother forms of causal inference (see, e.g.,Woodward 2002). Semiotic chains of action shift the scale of phenomena: instead of looking at phenomena athigh levels of abstraction andaggregation, they magnify the resolution of there searchers’ analytic anddescriptive lenses. But how can researchers convincetheir readers thatthe components of the process they emphasize areindeed the critical ones for understanding the phenomenonthey wish to explain?And, if researchers want to make a potentially generalizable point, how can theydistinguish incidental from regularly occurring causal processes?

Yet another challenge is to show how the different components of amechanism work together in a processual way. Peirce’s semiotics does not tellus why, in a specific case, meaning making operates as it does,only how tothink of meaning making’s constitutive parts. When people act indifferentsituations, they obviously react to what is being overtly said andseen (thesign-object) but they are also influenced and take into accountother, oftenmore difficult to directly perceive,aspects that are explicitly or implicitlyinvoked within the situation

—their histories, idiocultures (Fine1979), ideological determinations, and soon—what pragmatists termed their “habits ofthought andaction” (Peirce 1992; see also Gross 2009; Kilpinen 2009). The effectofthe sign-object in action, the interpretant, is thus always more than the sumofits overt “stimuli.” In essence, the challenge is still that of renderinganinherently “invisible” causal explanation visible. This is both because causality isnever directly evidenced (Hume [1740] 1967), and as the interpretant is neverdirectlyformed by the sign-object but also by characteristics of theinterpreter and therelationship between the actor and the situation.

Lui: 

(1) In Lecture 7, Aspectsof Sociological Explanation, I resolved the issue “scaleofphenomena” within H-D (Hypothetico-Deductivism) by specifying interaction=(hypothesis = structure-cum-society―definition of thesituation = data). The hypothesis can be on the macro-scale. 

(2) Like structure-cum-society,“histories, ideocultures,ideological determinations” are othernesses arising from agency in the ontology of our theoretical sociology. It is a littlestrange to term them “habits of thought and action”. 

(3) The notion of “the situation” used by Tavory and Timmermans maybe different from W. I. Thomas’s.Watch out when we read on.

P. 689:

We can, however, strengthen our mechanistic explanation togainmore confidence in a proposed causal account. Machamer et al. (2000, p.13)highlighted that mechanisms produce “regular” changes. To figure out what constitutes an accidental or a regular aspect of a chain ofcausation,ethnographers look for similarities and differences among semiotic chains.Ethnographers systematically examine action across observations,arguing that anobservation is “a case of . . .” a larger universe ofobservations that share asimilar causal structure (see Becker and Ragin 1992).As each instance is seenas one piece of evidence for a generalizable semioticchain, the very definitionof the ethnographic object recursively shifts (see also Lakatos 1976; Katz2001). This iterative process of redefinition providesinsight into the workingsof the proposed mechanism as it brings into purview salient resources and shared understandings in some cases but not in others. Ethnographersthus describe continuous and intelligible causal processes but also need tobuttress thestrength of their causal explanations by accounting for variation amongdifferent instances of a particularsemiotic operation.

Variation, as many have observed, depends on the notion of a“set”(see, e.g., Goertz and Mahoney 2012). Only when a common questionorcharacteristic already defines observations as comparable cases does it makesenseto compare them in the first place. But the very architecture of a set isrootedin the researcher’s theoretical assumptions. Here we develop three kinds of variation suggestive of causal explanations drawn from Peirce’s semiotics: data set variation looks at multiple instances of semiotic gestalts, variation intime examineschanges across a semiotic chain, and intersituational variation holds one aspect of the semiotic whole constant while examining itsconsequences in different cases.

Lui: 

(1) To account for variation among different instances is easierbut it may not imply a distinction between the accidental aspects from theregular ones. The latter task is much more difficult. 

(2)True, the three kinds of variation arise from the researcher’stheoreticalassumptions. 

 P. 690:

Data Set Variation

The first form of variation that is available to ethnographers, and the one that has received the most analytic attention, iswhat we term data setvariation. Here, theethnographer collects situations that seem—eitherprototheoretically or because of previous research—to be instances of “the samething.”

The broad semiotic set of sign-object-interpretantis assumed tobe constant; we observe more of the same situation. Having defined the setof situations, ethnographers proceedto sift through their similarities anddifferences that lead to various consequences, double-fitting observations andcausal explanations as they go—and often changing their assumptions about whatis, or is not “similar” asthey go along. Thus, although ethnography may be a “small N method” (Small2009), any ethnography contains a large number of similar cases withinthe purview of the study.The logic of data set variation shares much with Mill’s([1843] 2002) methodsof agreement and of difference to assess causality: interactionsand situations are compared in order to see how specific differences inthesituation lead to different outcomes. To theextent thatethnographers are interested in one kind of situation—be it“getting pissed off” while driving (Katz 1999),prison wardens typifying incoming prisoners as Black, White, or Hispanic(Goodman 2008), or how visibly disabled people manage interaction with “normals” (Davis 1961)—it isthedata set variation among similar cases that is crucial for the construction of the causal claim.

Although ethnographers seldom formalize such variations,this formof causality construction shares much with counterfactual logic andCharlesRagin’s (1987, 2000) qualitative comparative analysis (QCA). In thoseinstances,researchers compare observations following asubjunctive conditional, a what if logic.Whileeschewing counterfactual logic’s emphasis onnecessary conditions, ethnographers often “retrodict” from their observations,claiming that within a certain pattern of action, if X had been observed, then Ywould have happened (see Katz 2001). And, inethnography as in QCA, differentpatterns of meaning-making processes and [Lui:are?]compared and sufficient conditions aresought. Thus, whileethnographers are reluctant to assign a binary tag (Ragin1987) or a numericgrade to an observation as in later iterations of QCA (Ragin2000), theycontinuously ask, How similar is this situation to anothersituation in which aspecific outcome was evident?

Lui: 

(1) Tavory and Timmermans mention “interaction”, “situation” and“outcome”. With respect to the Piercian semiotics (sign―object)―interpretant(it iswritten as “sign-object-interpretant” in their paper, not quite right),sign =definition of the situation, object = the situation, interpretnat  =outcome of interaction. In my Lecture 7, Iconnect them as follows: interaction= (structure-cum-society―definition ofthesituation), incident = interaction, sequence of incidents. The first partisH-D, also a kind of explanation; the third part is social explanation. 

(2)Tavory and Timmermans do not mention the“structure-cum-society” (theothernesses). They give three examples of “interaction and definition”, namely,being insulted while driving, prisonwardens typifying incoming prisoners,interaction of the visible disabled with“normals”, all of which imply“structure-cum-society”. 

(3) The “outcome” of an incident is the one immediately followingit, and hence in social explanation the leading incidentis the cause of thefollowing one. Obviously, a leading more than one followingincident;vice-versa, a following incident can have more than one leading incident. Tavoryand Timmermans are talking about the latter kind. In the perspective ofmechanismic explanation, all these multiple pathways areincluded into themechanism, and hence it is not a problem of similarity but ofpathway multiplicity.I personally prefer the latter than the former. In thispreference, it is not aproblem of lean fact (data) but of fat fact (data-hypothesis).

(4) The “what-if logic” is I think just the usualhypothesis-testing by data instatistics, which can be applied to ethnographicdata in a conceptually generalized manner.  

P. 691:

Variation over Time

A second form of variation relates to meaning-makinginstances over time. The ethnographerbegins with the insight that meaning-making isan ongoing activity, where oneiteration of meaning making influences howmeaning is made next—where oneinterpretant becomes the sign of the next cycle.This is a form of variation particularly suited to the craft of ethnography.Ethnographers typically spend an extended time in the field andgather historicaldata alongside interactional observations. Ethnographers mayspend years withthe same people, observing how they make and remake sense oftheir worlds asthey move within a specific social career (Becker 1952; Strauss1993) or evenacross generations (see, e.g., Black 2010; Smith 2006).

To account for variation over time, the causal explanation mustexplain not only actions at a specific point of time but also transformations of meaning making. Like historical sociologists, ethnographersare thus interested in pathdependency of trajectories (Mahoney 2000), in the social structure ofturning points (Abbott 1997), and in tracing events overtime (Mahoney 2012).Relying on an assessment of salient differences and agreements, they assesscausality by examining how processes change over timeand across observations.

Variation over time highlights the limits of basing causal claimssolely on data set variation. Most ethnographic causalexplanationssimultaneously account for data set variation of situations at eachpoint intime and for change over time. The resulting set of observations,however, isnot a qualitative “panel study” in which data set variation in two points oftime are compared. This is not only because ethnographers attempt to follow actors and meaning making more or less continuously, but also because the kinds of situations that they observe as time goes by change as well. For example,thechallenges that a social movement faces in its first months are different than those one year later when it must decide whether to institutionalize aspecificidentity (Blee 2012). Religious converts feel the presence of thedivine differently,and in different situations, as they settle into their new religious lives (Tavoryand Winchester 2012). Looking at action over longertemporal arcs thus pushes ethnographers to look at what may seem like “apples and oranges” to the quantitative methodologist.

Lui: 

(1) It is clear that “variation over time” is just that a leading incident with more than one following incident. In fact,data set variation andvariation over time are captured very adequately by thenotion of mechanism. Itmay be an indication that Tavory and Timmermans haves hallow reading of the mechanismic explanation. Watch out when you read thelater section FromVariation to Mechanism, which I shall not comment on. 

(2)The coverage provided the it can be easily extended by varyingthe explanans(in the natural science the phenomenon is the set of all observables from whichone or more of them are selected as the explanans to be explained by the mechanism as its explanandum) to generate more mechanisms.

P. 692:

Intersituational Variation

Intersituational variation constitutes the callingcard of ethnographic research, a form of variation thatis harder to constructin other methodologies. Here, the researcher collects actors’—and notnecessarily the same actors’—actions in different settings and situations andshows that seemingly unrelated actions make sense as a singleset under theresearcher’s theoretical description. In other words, the researcher keeps one aspect of the meaning-making process constant to examine how different situations are refracted through—or transformed by—these semiotic aspects. Thus, for example, researchers canfollow an object across situations(see, e.g., Tsing 2005), or look at howactors’ habits of thought and actionstructure different situations they mayencounter. As opposed to data setvariation, the search for intersituational variation partly rests on theprototheoretical understanding that thesituations the researcher collects intoa set are not similar; and, as opposedto variation over time, the ethnographeralso does not assume that meaning makingin one situation structures meaningmaking in others. Rather, the situationsare made comparable through the ways inwhich a shared characteristic affects atheoretically constructed causalexplanation. Most paradigmatic urban ethnographies—from Whyte’s (1943) StreetCorner Society to Liebow’s (1967) TallyCorner and Duneier’s (1999)Sidewalk—describe their protagonists as theynavigate different kinds ofsituations.

A causal account emerges largely through the ways in which the ethnographer accounts for different interactional outcomes that ariseindifferent situations. Thus, one of the most important observations inLiebow’s analysisof the fractured lives of inner-city black men is the ways inwhich theynegotiated relationships differently with other men “on the corner”and withtheir female partners. The failed attempt to live up to a standard of masculinity gains explanatory breadth when we see the men bragging about theirsexual prowess and manipulative attitude toward women among their peersandwhen we simultaneously witness their relationships fall apart because theyfeelthey cannot provide for their female partners. Positing the attemptstogenerate a competent masculine persona in the face of structuraldiscriminationas his causal engine, Liebow can thus account not only forsimilarities amongthe actions of different people on the corner over time andemploymentsituations, but also for how they make sense of their lives acrosssituations,even if their actions may seem to be, at first, contradictory.

Except for relatively straightforward cases—where, forexample,class position (e.g., Bourdieu 1984) or personality structure (e.g.,Stouffer1955) are already supposed to affect a range of tastesor activities—ethnographic intersituational variation is difficult toformalize,as it groups heterogeneous situations. The ethnography-specific causaltheorization is precisely what makes the different situations and observations comparable. If we argue, for example, that there is a distinct“code of thestreet” (Anderson 1999) that governs inner-city black men’sactions, we expectto see such a code in action, among other situations, inconfrontations on thestreet, in job interviews, and in ways of seeing imaginaryslights in mundane interactions This does not mean, however, that men wouldbehave similarly inthese diverse situations, but that it is precisely the diversity of the forms and outcomes of interactions in different settings thatthe causal claim should explain. As opposed to the variation between obviously similar situations andactions, intersituational variation makes sense onlyunder a specific causaldescription.

Defining and exploring variation is central to developinganethnographic causal account. Through variation, ethnographers isolateanalytically important elements in actors’ meaning-making process within theircase.Ethnographers also come to recursively define the case by linking itthrough recurring elements—both discovering an account and justifying it.Still, evenlayering a mechanism-based semiotic approach and subjecting it tothe tests ofvariation is insufficient for a convincing causal explanation. Whilesemiotic specification and variation may increase our confidence in a causalclaim,there are inevitably multiple causal explanations possible forconfigurationsof observed variation. How can researchers sort thesealternatives? The question, then, becomes how to evaluate the ethnographiccausal account’s explanatory fit and how to judge this account in relation toother plausible alternatives.

Lui: 

This long section boils down to one pointwhich Tavory andTimmermans perhaps miss, that is, they have not considered H-Din order to takeothernesses (structure-cum-society) on board. Theintersituational variation isa matter of course if otherness is taken intoconsideration because everyinteraction demands some particular otherness thatmakes the situation intelligibleand hence definable. I stop my commenting at this point, and the remaining isleft to your own reading. 

————————

（Sociological理论大缸第62期）

超链接：

第60期 百年《美国社会学学报》,只有54篇“历史”论文？！清单。

第57期 “我们需要的不仅是机制！”—阐释论的不满与反攻

第56期 民族志也能【做出】机制式因果解释？实用主义回答

第55期  别太抽象理论、别太微观行动，快告诉我【机制】到底要怎么找！

第51期 活着的美国社会学理论家有三代？关注哪些议题？《ST》主编Emirbayer透露