mobilization and counter-mobilization processes: from the “red years” (1919–20) to the...

Mobilization and counter-mobilization processes:

From the ``red years'' (1919^20) to the ``black years''

(1921^22) in Italy

A new methodological approach to the study of narrative data

ROBERTO FRANZOSI

Trinity College, University of Oxford

``One of the consequences of the introduction of quantitative methods

[in history] has been the discovery that church records, probate records,

tax rolls, and similar sources contain detailed information on a wide

variety of human activities stretching far back into time.''

1

These words

by Fogel correctly point to the bustling activity of quantitative historians

in this century as they spared no e¡ort in their attempt to produce

quantitative evidence. But there is another potential source of quantita-

tive evidence that Fogel does not mention: narrative itself, as produced

in copious quantities by chroniclers, analysts, policemen, and journal-

ists; their reports, too, provide a wealth of information on human

activities. Needless to say, modern historians have made ample use of

both police and newspaper reports. Spriano

2

and Maione,

3

just to men-

tion two leading Italian historians who produced some of the seminal

work on the period studied in this article, relied mostly on police records

and newspapers for evidence. I have argued, however, that this abundant

information found in narrative form and typically also used by histor-

ians in narrative form can be used in more systematic ways within a

quantitative framework.

4

On this score, social scientists have been gen-

erally more daring than historians, producing a number of important

studies based on newspapers as sources of historical data.

5

Within his-

tory itself, perhaps Tilly alone has made systematic use of police and

newspaper narratives in quantitative fashion.

6

In this article, following Tilly's footsteps,

7

I develop a formal method

based on linguistics for the collection of narrative text. I discuss both the

advantages and disadvantages that come with the adoption of such a

formal approach to content analysis. I illustrate the power of the tech-

nique by providing some exploratory analyses of data that I collected

from an Italian newspaper on one of the most turbulent periods of

KAP ^ RYSO 399 ^ PIPS Nr. 125454 Rapati 9-5-97 09:48 Page 275

Theory and Society 26: 275^304, 1997.

ß 1997Kluwer Academic Publishers. Printed in the Netherlands.

Italian history, between 1919 and 1922, from the end of the First World

War to the rise of Fascism. The method should be of particular interest

to historians and social scientists who hold a view of history as agency

and process, i.e., a view that history unfolds through the actions of social

actors organized in the pursuit of their interests and with social inter-

action among those actors providing the dynamic for social change.

After all, the method has at its core the set of actions performed in time

and space by a set of social actors.

A formal approach to content analysis based on linguistics

The problem of analyzing text data in statistical ways (that is, the prob-

lem of going from words to numbers) has fallen under the heading of

content analysis. Typically, content analysis has solved the problem via

the use of a coding scheme, a set of categories designed to capture the

input text. Unfortunately, content analysis is plagued by a variety of

shortcomings. To overcome these shortcomings, I propose an approach

to the analysis of text data that radically departs from traditional con-

tent analysis. The approach is based on a highly formalized semantic

grammar.

Semantic grammars have a long tradition. Their origin can be traced

back to the work of the Russian morphologist Vladimir Propp. In his

study of the semantic structure of Russian folktale,

8

Propp found that

thirty-one functions capture the basic, deep, underlying structure of all

Russian folktales. The functions basically correspond to the actions that

the hero of the story performs. Greimas

9

later reduced Propp's thirty-

one functions to ¢ve. These schemes narrowly focus on actions, rather

than actors, presumably because the hero is the protagonist of the story.

Labov and Waletzky

10

more generally showed that narrative stories are

characterized by both actors and actions in a time/space dimension.The

very surface structure of narrative stories (i.e., the way they are syntacti-

cally organized) is typically patterned after the basic canonical form

of the language: subject, action, object, and their modi¢ers. Similarly,

Halliday

11

talked about sentence structures in terms of: 1) actors, 2)

processes (or actions) and related instruments, 3) goal or bene¢ciary

(the object), and 4) time and space. Rumelhart

12

and Mandler

13

argued

that these story schema can be described by agrammar, a story gram-

mar, similar to a sentence grammar. But, while a sentence grammar

speci¢es a story in terms of syntactic categories, a story grammar speci-

¢es the same story in terms of functionally relevant, semantic categories.

KAP ^ RYSO 399 ^ PIPS Nr. 125454 Rapati 9-5-97 09:48 76

276

In particular, a story grammar 1) speci¢es a set of functional categories

and 2) provides the rewrite rules that ``specify the linear and hierarchial

ordering of the categories in a `well-formed' narrative structure.''

14

Following Labov andWaletzky

15

and Halliday,

16

the grammar I propose

organizes sets of actions performed at a particular place and time by

some actor(s) against or in favor of some other actor(s) into events, and

events into higher-level aggregates, such as disputes. All of this can be

formally represented in a grammar through a series of rewrite rules.

dispute y {5event4}

5event4 y {5semantic triplet4}

5semantic triplet4 y {5subject4} {5action4} [{5object4}]

These few lines show that the initial symbol

17

dispute can be rewritten

(y) as a set of one or more (as indicated by the curly brackets) events;

the non-terminal symbol 5event4 (non-terminal as it is enclosed in

angular brackets) can be rewritten as one or more sets of5semantic

triplet4; a 5semantic triplet4 is the combination of one or more

subjects, actions, and objects, where the square brackets [ ] around the

object make the object optional (e.g., ``employers lockout'' or ``workers

strike'').

Subjects, actions, and objects can be further rewritten to capture sub-

stantively relevant attributes of actors and actions. For instance, one may

want to know the number of individuals involved; the name, profession,

rank, marital status, age of an individual; the demographic character-

istics (young, old, male, female, black, white) and the organizational

a¤liation of individuals and groups. Organizations operate in di¡erent

societal spheres (religious, political, social, economic; economic organ-

izations operate in di¡erent sectors: chemical, textile, banking, etc.), at

di¡erent levels of aggregation (e.g., shop, plant, ¢rm). Thus, we can re-

write a5subject4 in the following way:

5subject4 y 5actor4 [{5actor modi¢er4}]

5actor4 y (management Aworkers A unions A government A ... )

5actor modi¢ers4 y [{5number4} {5type4}] [{5organization4}]

5number4 y (one A two, ... , tens A hundreds, .. . , some A several, . .. )

5type4 y (female Ayoung A Hispanic A skilled A .. . )

5organization4 y (Ford A United Airlines A United States Government A .. . )


277

Actions may be similarly characterized by several modi¢ers. An action

occurs in time and space. An action has a time duration and an out-

come(s). An action, e.g., ``strike,'' can be of several di¡erent types, such

as ``general,'' ``wildcat,'' ``sit-in,'' or ``checkerboard.'' The action may be

based on di¡erent instruments. For example, a union leader may address

a group of workers with a ``loudspeaker''; or an employer may notify a

union of an imminent round of layo¡s via a ``telegram.'' A group may

express di¡erent grievances ^ ``economic,'' ``political,'' ``the abdication

of the king.'' In summary, the non-terminal symbol5action4 may be

rewritten as follows:

5action4 y 5action phrase4 [{5action modi¢er4}]

5action phrase4 y (strike A demonstrate A riot A .. . )

5action modi¢ers4 y [{5type}] [{5instrument4}] [{5reason4}] [{5outcome4}]

[{5space4}] [{5time4}] [{5duration4}]

5type4 y (general A sit-down A .. . )

5instrument4 y (baton A tear gas A telegram A loudspeakers A ... )

5reason4 y (wages Aworking conditions A .. . )

5outcome4 y (positive A negative Awage increase A ... )

5space4 y (Turkey A Greece A ... A Genoa A NewYork A . .. )

5time4 y (March 15, 1453 A .. . A three years ago A .. . 2 a.m. A in the

afternoon A .. . )

5duration4 y (8 hours A several days A inde¢nite A . .. )

Finally, the grammar rewrites the5object4 as a subject (which basi-

cally is a human actor or an institution) or an inanimate noun (e.g.,

factory, road). This way expressions such as ``workers occupy the fac-

tory,'' ``protesters set ¢re to cars,'' and so on, can easily be coded in the

grammar.

5object4 y 5subject4 A5inanimate object4

5inanimate object4y (factory A road A car,... )

An example: From the ``red'' years to the rise of Fascism (1919^22)

In the course of 1919, discontent was on the rise throughout Italy. Indus-

trial employment was sluggish and prices were soaring. War veterans

were increasingly disgruntled over the fact that the government was slow

to deliver on its promises of rewards (e.g., land reform) for the veterans'

contribution to the war e¡ort. Mass rallies againstcarovita (in£ation) in

the cities and peasants' occupations of uncultivated lands and estates in

the countryside became common. Industrial strikes soared. A few iso-


278

lated instances of factory occupations occurred here and there in early

1920.

18

For instance, on June 5, 1920, an article fromIl Lavoro, a socialist

paper from Genoa reads:

Brescia, 4th evening. Factory occupied by workers.We have news from Lovere

that workers at the Bianchi-Gregorini ¢rm occupied the plant, following

management's refusal to accept the workers' wage demands. It appears that

the plant manager was wounded...'' (Il Lavoro, June 5, 1920, page 4, column 2).

Isolated cases of this kind punctuated at ¢rst the rhythm of a harsh

confrontation between labor and capital. But it is the dispute over the

metalworkers' contract renewal, involving the largest and strongest

industrial union, that brought Italy on the verge of a revolutionary

confrontation. The negotiations over the renewal of the collective con-

tract of the metallurgic workers had been dragging on for months during

1919 and 1920 without appreciable breakthroughs toward a solution. By

1920, spirits were running high on both sides with rising numbers of

episodes of violence. Finally, on August 31, 1920, one reads inIl Lavoro:

The struggle of metallurgical workers.The lockout of aMilanese ¢rm provokes

factory occupation. Milan, August 30, evening. Following yesterday's inci-

dents, the Romeo company announced a lockout. Naturally, the metallurgical

workers of other plants in the area did not accept this decision passively and

the leaders of the labor movement immediately decided, in protest, that all

workers should occupy their factories. The order was issued and immediately

executed. In all the Milanese plants involved in the metallurgic workers dis-

pute, the workers did not leave their factories, and ... they are still in there,

without doing any work. At ¢rst, all technical personnel, the factory managers,

and some of the owners were kept inside the plant by the workers. In the

evening, however, almost all of them were freed. In the evening, in many

plants, workers' families brought in dinner. All in all, no serious incidents have

occurred... (Il Lavoro, August 31, 1920, p. 3, c. 3).

That day, August 31, and that factory, the Romeo works in Milan, mark

the beginning of the factory-occupation movement, the most salient

event in a two-year period known in Italian history as thebiennio rosso,

the red years.Within a day or two, at the beginning of September, most

metallurgic factories throughout the country came under workers' con-

trol. The following newspaper articles from Il Lavoro detail the begin-

ning of that momentous event.

New phase of the metallurgical workers' dispute. The factory occupation

widens... . In Rome. Rome, 1 evening. The metalworkers' dispute unfolds in

Rome without noteworthy incidents ... in the factories occupied by the workers


279

work continues in an almost normal fashion.... InTurin.Turin, 1. At a meeting

of metallurgical employers that began at 2 am last night, employers decided to

lockout their factories starting this morning. In fact, when workers came to

work at their usual time, they found the factory gates closed. The workers,

however, convinced the custodians to let them in and occupied the plants.

FIOM's appeal. FIOM [the metalworkers' union] released the following appeal

to the workers: ``. .. [employers] wanted to test the terrain with a lockout in one

of the most important Milanese factories, and they got the response that they

deserved. To the lockout of the Romeo ¢rm, the Milanese Strike Committee,

acting in accordance to the guidelines previously decided upon and to the

orders issued by the Federal Committee and by the General Strike Committee,

responded with an order of occupation of Milanese ¢rms... . Italian metal-

lurgic workers ... get ready to follow the instructions of your organization. In

the areas where employers will attempt to lockout you will be asked to follow

the example of your Milanese comrades. Long live working class solidarity!

The Federal Committee, The Strike Committee (Il Lavoro, September 2, 1920,

p. 2, c. 1).

Starting September 3, 1920, Il Lavoro began reporting on the factory

occupation on the second page with titles running the width of the page:

the scale and magnitude of the event, its implications for the social

structure of the country did not escape the contemporaries. For Lyttle-

ton, it is appropriate to consider the situation as revolutionary because

the actors perceived revolution as probable.

19

After all, new reports

about events in Russia were making headlines every day.

The metallurgic workers' dispute in Italy. Workers occupy their factories.

FIOM's directives. September 3. The regional Ligurian Committee of the

Italian Federation of metallurgic workers communicates: yesterday morning,

the metallurgic workers of the entire Liguria region went to work as usual, but

found the factory gates closed with the police standing guard. Employers,

without any announcement ... had started a lockout. After the ¢rst initial

surprise, all the workers went in to work ... and occupied the factories .. .. In

Naples. This morning the metallurgical workers occupied the Ilva works... . In

Livorno. Today at 4 pm the metallurgical workers occupied the local plants.

In Bergamo. Following the lockout called by the Milanese metalworking

employers ^ even though the Bergamo employers had not locked out ^ the

workers of some factories occupied their plants today in the afternoon in the

city and in the province. In some cases, workers kept the technical personnel

and accountants inside the plants... (Il Lavoro, September 3, 1920, p. 2).

During the crucial days of the factory occupation movement throughout

the month of September 1920 the Giolitti government adopted a hands-

o¡ approach to social con£ict. Giolitti's plan was clever: wait for the

movement to de£ate and reap the political bene¢ts of success. Un-

fortunately, the plan failed. The movement did de£ate, but the govern-

ment posture enraged the business class who had felt unprotected under


280

threat. The mood of the middle and of the upper classes swung to the

Right as the ``red scare'' swept the country. By late 1920, no sooner had

the working class mounted its largest mobilization in Italian history,

than episodes of Fascist violence started to rise in frequency in the

country side at ¢rst, and by 1921 in the cities as well. In 1922, a handful

of Fascists ``marched on Rome'' and took power. It would be another 20

years before the Italian working class regained its freedom, another ¢fty

years, during the 1969 ``hot autumn,'' before it could mount a mobiliza-

tion process of the scale of thered years.

20

Coding one of the newspaper articles within the categories of a semantic

grammar would yield the following code:

5dispute4 metalworkers

5event 14 incidents5time4 yesterday

5event 24 5semantic triplet 14 5subject4 ¢rm (5name4 Romeo)5action4

announce lockout

5event 34 factory occupation (5comment4 no serious incidents have occurred)

5semantic triplet 14 5subject4 leaders (5organization4 labor movement)

5action4 decide factory occupation (5reason4 protest)

5semantic triplet 24 5subject4 workers (5economic sector4 metalworking

5number4 all factories)5action4 do not leave plant

5semantic triplet 24 5subject4 workers5action4 do not work

5semantic triplet 34 5subject4 workers5action4 keep inside the factory

5object4 technical personnel5object4 managers5object4 owners5number4

some)

5semantic triplet 44 5subject4 workers5action4 free (5time4 evening)

5object4 technical personnel (5number4 almost all)5object4 managers

(5number4 almost all)5object4 owners (5number4 almost all)

5semantic triplet 54 5subject4 families5action4 bring dinner (5time4 evening

5space4 most factories)5object4 workers

A set theoretical representation of a semantic grammar

The example of coding provided above certainly makes it seems that a

semantic grammar can ``capture'' narrative text. But we are still far from

having gone ``from words to numbers.'' What we have done is simply

take words and structure them within the organizational framework of

a semantic grammar. But, we are still in the realm of words. How do we

go from there to the numbers? Furthermore, how can we analyze the

complex set of relationships between subjects and actions, actions and

objects that the grammar makes explicit? The answer to those questions

lies in the mathematical underpinnings of the formal representation of

a semantic grammar, namely set theory. The concepts that set theory


281

makes available, such as set, cardinal number, ordered pair, Cartesian

product, and relation, make possible the quantitative analysis of words

and of their interconnections. Let me brie£y de¢ne these concepts.

A set is nothing but a collection of objects of whichever kind: cars in a

parking lot, children in a school, oranges in a fruit bin. More to the point,

in the example of a newspaper article coded above, the5subjects4

{¢rm, leaders, workers, workers, workers, workers, families} constitute a

set; and so do the5actions4 {announce lockout, decide factory occu-

pation, do not leave plant, do not work, keep inside the factory, free,

bring dinner}. In a notation called theset-builder form of a set, the set of

5subjects4 S can be expressed as S = {s A s is subject}, which reads ``S

is the set of subjects s such that (the symbol A reads `such that') s is a

subject.'' The notations s [ S and s N S read respectively ``s belongs to S''

and ``s does not belong to S.'' Thus, the element ``workers'' belongs to the

set of5subjects4 but the element ``police'' does not.Thecardinal num-

ber of a set measures the size of a set; basically, it gives the number of

elements in a set. Referring back to the coded example above, the cardi-

nal number of the set of subjects is 7 and that of the set of actions is also

7. The cardinal number of the subset of distinct subjects is 4 because the

same subject ``workers'' appears 4 times, while that of the actions is still

7, as no action is repeated. It is the cardinal number that allows us to

reexpress a list of words in numeric terms. Similarly, Cartesian product,

sentence, and relation express the links among the categories of the

grammar in a tight mathematical formulation.

The product set orCartesian product of any two sets A and B consists of

all ordered pairs (a,b) where a [ A and b [ B. It is denoted by A ¾ B

(reads ``A cross B''). Using the set-builder notation, this isA ¾ B = {(a,b)

A a [ A, b [ B}. Thus, if S is the set of subjects S = {¢rm, leaders,

workers... } and A is the set of actions A = {announce lockout, decide

factory occupation... } the Cartesian product of S and A consists of all

ordered pairs (¢rm, announce lockout) (¢rm, decide factory occupa-

tion) ... (workers, announce lockout) ... More generally, the Cartesian

product of n sets S1

, S2

, ... ,Sn

is denoted by S1

¾ S2

¾ S3

¾ ... ¾ Sn

and

consists of all n-tuples (s1

, s2

, s3

, .. . , sn

) where s1

[ S1

, s2

[ S2

, ... , sn

[ Sn

.

Thus, a semantic triplet is a Cartesian product S ¾ A ¾ O, where the set

O would contain a null element to account for triplets with no object

(e.g., ``workers strike,'' as opposed to ``workers hurl rocks at the police'').

Needless to say, only a subset of the 2-tuples that result from the Carte-

sian product S ¾ A ¾ O will be semantically well-formed. Thus, the


282

2-tuples (¢rm, decide factory occupation) or (workers, announce lock-

out) do not make sense in light of capitalist social relations of produc-

tions. Under those social relations, ¢rms lockout and workers occupy

factories but not vice versa. The concepts of sentence and relation allow

us to restrict to semantically acceptable n-tuples the combinations that

result from joining sets. Apropositional function or sentence, P(a,b), has

the property that P(a,b) is either true or false for any ordered pair (a,b)[

A ¾ B. The relation R = (A,B,P(a,b)) is the Cartesian product of sets A

and B subject to the constraint of the sentence P(a,b). If R = (A,B,P(a,b))

is a relation, the solution set R*of the relation R consists of the elements

(a,b) in the Cartesian product A ¾ B for which P(a,b) is true. Notation-

ally, R*= {(a,b) A a [ A, b [ B, P(a,b) is true}. Thus, the set of semantic

triplets of the grammar is the subset of acceptable 3-tuples obtained

from the Cartesian product of the sets of subjects, actions, and objects

S ¾ A ¾ O, i.e. T = {SX AXO A (s,a,o) N C, s [ S, a [ A, o [ O} where

X is the set symbol for a union or sum of sets (in this case, of the sets S,

A, and O) and C is the set of ``selection restrictions,'' i.e., the set of

invalid 3-tuples.

The purpose of expressing a semantic grammar in a set-theoretical nota-

tion, rather than in a linguistic notation, is not simply one of ``form'' and

notational elegance.The purpose, rather, is that of being able to use tools

available in one notational domain but not in the other. Indeed, such

concepts as cardinal number, Cartesian product, sentence, and relation

within the set-theoretical domain provide the basic tools for going ``from

words to numbers'' and for expressing mathematically the complex

relations between words.That, of course, still does not solve the practical

problem of just how one goes ``from words to numbers.'' As it turns out,

however, set theory serves as the mathematical foundation of relational

data base systems (RDBMS). RDBMS provide powerful models of com-

puter data organization and storage. It is precisely because a grammar of

data collection can be implemented in a computer environment through

relational-data-base technology that content analysis based on such com-

plex coding schemes as semantic grammars becomes practically feas-

able.

21

Let us explore what relational data models have to o¡er for the

practical implementation of a semantic grammar in the computer, and

the role that set theory plays in bridging the gap between semantic

grammars and relational data models.

Key to relational data models is the organization of information in

separate but interconnected ¢les (ortables, as ¢les are called in relational

vocabulary), rather than in a single rectangular ¢le. Together, the set of


283

tables constitutes a data base. Relations between tables are established

via the presence of at least one overlapping ¢eld (or column) across

tables. Thus, for instance, the following two tables (actor�table and

action�table) that contain the basic information of the semantic triplets

of the coded example, have one column in common, the triplet�id (see

Table 1). The same triplet�id in the two tables tell us that the corre-

sponding actor�name(s) and action�name(s) go together. The column

``object�id'' in the actor�table identi¢es the function of the actor�name

as either a subject or an object in the triplet (0=subject; 1=object). For

example, the triplet�id = 5 in the actor�table points to the actor�name

``workers'' as subject (object�id=0) and to the actor�names ``technical

personnel, managers, owners'' as objects (object�id=1) and to the

action�name ``keep inside the factory'' as action in the action�table,

which corresponds to the properly formed English sentence ``workers

keep technical personnel, managers, and owners inside the factory.''

In summary, via the triplet�id it is possible to reconstruct the original

narrative of the article. All we need is a general way to search and

extract information from (``querying'') the data base. SQL (Structured

Query Language) provides just such a general tool with a simple set of

commands (namely, select, from, where). Both single- and multiple-table

queries are possible in SQL. For example, the following single-table

query operates on the actor�table.

Table 1. Relational tables (actor�table and action�table)

actor�table action�table

triplet�id object�id actor�name triplet action�name

1

2

3

4

5

5

5

5

6

6

6

6

7

7

0

0

0

0

0

1

1

1

0

1

1

1

0

1

¢rm

leaders

workers

workers

workers

technical personnel

managers

owners

workers

technical personnel

managers

owners

families

workers

1

2

3

4

5

6

7

announce lockout

decide factory occupation

do not leave plant

do not work

keep inside the factory

free

bring dinner


284

select actor�name

from actor�table

where triplet�id = 5 and object�id = 1

and yields the list of actor�names as ``objects'' for triplet�id equal to

¢ve:

actor name

technical personnel

managers

owners

The following multiple-table query operates on both the actor�table and

the action�table

select actor�name in actor�table and action�name in action�table

from actor�table and action�table

where ((triplet�id in actor�table = 1) and (triplet�id in action�table = 1) and

(triplet�id in actor�table = triplet�id in action�table))

and yields the ¢rst stripped-down triplet of our narrative (i.e., a triplet

without modi¢ers):

actor name action name

¢rm announce lockout

Gains and losses of formalism

I have shown elsewhere that a grammar of data collection presents

several advantages over more traditional approaches to content analy-

sis.

22

To summarize brie£y: 1) Semantically de¢ned and highly dis-

aggregated, functional categories (rather than abstract, general, and

theoretically de¢ned categories) provide the basic building blocks of the

grammar. 2) The coding categories are fully relational, and the relations

are explicitely de¢ned by the grammar: subjects are related to actions,

actions to objects, and all to their respective modi¢ers. 3) The grammar,

as a coding scheme, limits the substantive and theoretical role of the

coder, while making explicit her or hislinguistic role in text understand-

ing, in mediating between source material and coded output. 4) Coded

output from a grammar is highly disaggregated; output preserves much


285

of the linguistic £avor of the input material, both in terms of syntactic

structure and vocabulary. Given that all relevant information is coded in

ways that closely resemble the original text input, data collection is as

hypothesis-free as possible. Thus, the data are highly re-usable. 5) A

grammar provides data in a format (actors/actions) that should be of

particular interest to those historians and social scientists who view

history as agency and process.

23

In this view, history unfolds through

the actions of social actors: social change is the result of a constant

process of social interaction among historical actors.

These are no small gains from a method of collection of historical data.

Yet, the greatest advantage of a linguistic approach to content analysis

derives from the fact that a semantic grammar easily translates into a set

theoretical framework. There are several advantages in moving from a

linguistic to a mathematical notational framework. First, set theory pro-

vides a simple mechanism to go ``from words to numbers'': the cardinal

number.When we move from words to numbers we move into the more

familiar terrain of statistical data analysis. Second, set theory provides a

powerful mathematical notation (namely, Cartesian product, sentence,

relation) for expressing the complex set of relations between the building

blocks of grammar (the old ``coding categories''). Third, set theory con-

stitutes the mathematical foundation of Relational DBMS methodology.

It is this methodology that makes feasible the practical use of such

complex coding schemes as semantic grammars in a computer environ-

ment.

Yet, for all its power, the linguistic approach to content analysis that I

have proposed has its limitations. The gains from formalism have come

at the expense of universal applicability. The advantage of traditional

content analysis is that, precisely because of itsad-hoc application, it can

be used for any kind of input text.The researcher will tailor the design of

the content-analysis scheme and of the coding categories to both the text

and the theoretical framework. A semantic grammar, on the other hand,

can only be applied to input text that is syntactically organized around

the canonical form of the language: noun phrase/verb phrase, or, sub-

ject, action, object, and modi¢ers as described here. That requirement

basically restricts the type of input text for which a grammar would be

applicable to narrative text, i.e., text organized along the time and space

dimensions, with actors performing certain actions. This broadly corre-

sponds to the ¢veW's of journalism:who, what, where, when, and why.

24

Unfortunately, not all text genres fall under the heading of narrative. For

example, scienti¢c journal articles, philosophical treatises, and even parts


286

of discourse found in narrative texts do not have the linguistic properties

of narrative text. Fortunately, however, many types of documents for

socio-historical research do belong to the narrative genre (e.g., police or

newspaper accounts of collective actions).With thiscaveat in mind, that

a linguistic approach to content analysis works well with inputnarrative

text, a linguistic approach to content analysis provides a powerful and

highly formalized methodological tool for sociohistorical research.

Navigating the 1919^22 data base

It is perhaps an easy game to show the power of a technique by coding a

handful of contrived and carefully chosen examples. But the proof of the

pudding is in the eating. Does the power of the technique stand up

against the test of real data? To answer that question let me turn to some

simple exploratory analyses of data on over three thousand disputes that

I collected from over 15,000 articles taken from the Italian newspaper

Il Lavoro, of the 1919^22 period, as shown in Table 2.

Needless to say, I did not code 15,146 newspaper articles for the sake

of testing the power of a technique, however innovative. In fact, I

proceeded the other way around.

25

My venture into methodological

development was the result of a theoretical (or, more modestly, substan-

tive) quest. Brie£y, my work on Italian class con£ict on the basis of

o¤cial strike statistics had led me to the following conclusions.

26

First,

to understand what workers do (e.g., strike), we need to understand what

employers and the state do, or, as I have come to put it, ``if it takes two

to tango, it takes at least two to ¢ght.'' Second, strike waves, i.e., the peak

moments in working-class mobilization, such as the1919^22 and 1968^

72 waves in Italy, represent crucial turning points in the relations be-

tween labor and capital.Unfortunately, o¤cial strike statistics,with their

emphasis on highly aggregated ¢gures on numbers of strikes, strikers,

and days lost, can hardly help in addressing these claims empirically.We

need much more disaggregated information on the various actors in-

volved and on their respective actions. It is the empirical prescriptions of

Table 2. Frequency distribution of articles and disputes in the database

1919 1920 1921 1922 Total

Number of articles

Number of disputes

4213

597

4329

622

3715

1132

2889

926

15146

3277


287

those two substantive claims that sent me on a search for appropriate

methodologies.What I found were semantic grammars.

The years between 1919 and 1922 mark a fascinating period of Italian

history. In the brief span of just four years, we go from the height of

working-class mobilization to the height of elite counter-mobilization.

The ¢rst two years (1919^20) are known as thered years for their intense

social and labor unrest culminating in the factory occupation movement

of September 1920; the last two years (1921^22), also known as theblack

years, saw a rapid increase in Fascist violence ultimately culminating in

Mussolini's march on Rome in October 28, 1922.

Querying the 1919^22 database with its many short and long narratives

on thousands of events should allow us to identify common properties

across the di¡erent events, temporal and spatial trends in those proper-

ties, networks of interaction, points of historical structural break and

change: event and structure, as Braudel would have it, built into the rich

texture of the data.

Table 3 shows the frequency distribution of the most common actors in

the database. The data clearly show a shift from a prevalence of work-

ing-class actors in the early years (the red years 1919^20; e.g., trade

unions, workers) to a prevalence of capitalist-class actors during the

following black years (employers, in particular). Remarkable, also, is the

increasing presence of political activists during the later years. The

government is notably absent during the crucial years of 1920^21, during

the factory-occupation movement, and the aftermath, during the ¢rst

wave of backlash from the right. That, indeed, had been President

Giolitti's strategy against the Left and the Right, a strategy that turned

Table 3. Frequency distribution of most common actors (number of disputes)

Types of actors 1919 1920 1921 1922

Trade unions

Employers

Workers

Government

Parliament

Ministries

Prefect

Police

Political activists

215

133

337

71

26

78

47

32

12

175

155

440

56

59

65

50

108

126

160

97

385

42

92

57

27

106

652

176

239

92

70

66

40

32

74

759


288

out to be politically very costly to both Giolitti and the country. The

Prefects (the local representatives of the government) and the police,

however, were very present during the years of maximum working-class

mobilization. This simple frequency distribution of actors thus seems to

highlight the process of mobilization from below during the red years

and counter-mobilization process from above during theblack years and

to underscore a shift in class power from the Left to the Right during the

same period.

The counts of Table 4 on the types of actions basically support the story

that Table 3 tells: a widespread mobilization process where claim mak-

ing, collective actions, and labor contracts prevail in the early years

followed by a sharp decline in working-class activism and a sharp in-

crease in political activism and violence during theblack years. In going

``from words to numbers,'' we have thus uncovered historical patterns,

we have provided empirical evidence on processes of theoretical interest,

such as those of social mobilization and counter-mobilization. It will

take much more work of analysis of these hard-won data, but the

beginnings are encouraging. Ultimately, it is results such as those of

Tables 3 and 4 and of subsequent Tables that make social scientists'

involvement in methodological work worthwhile. In Le Roy Ladurie's

words: ``The computer ... [will only tell us] what everyone knew in the

¢rst place....What matters is not so much the machine as the questions

one asks it.''

27

Again, the data of Table 5 on the distributions of demands con¢rm the

view of a working class on the o¡ensive during the ¢rst two years against

both the state and employers (e.g., 307 disputes with economic demands

in 1919 vs. 80 of them in 1922, but also demands for improved working

conditions, trade-union rights, contract renewals). The number of dis-

putes characterized by o¡ensive demands sharply dropped in 1921 and

1922 while defensive disputes (e.g., against layo¡s) increased. The num-

Table 4. Frequency distribution of most common actions (number of disputes)

Types of actions 1919 1920 1921 1922

Claim making

Collective actions

Meetings

Labor contracts

Verbal critiques and accusations

Violence

196

335

236

142

21

47

91

328

221

111

29

168

37

228

186

99

121

550

4

249

131

35

22

525


289

ber of defensive disputes against political violence increased but not

nearly in proportion to the increase in the frequency of violent actors

and actions. The Italian working class seems to have even lost its ability

to protest defensively by the end of 1922.

For all the clarity of the message of Tables 3 through 5, the high level of

aggregation of the data does not do justice to the mobilization process

that characterized the 1919^20 red years. A more ¢ne-grained look at the

information available in the data base would reveal that mobilization

took on many forms and involved a variety of social actors during those

years: From women to students, from war veterans to the unemployed,

from agricultural laborers to industrial workers ^ all took to the streets

in a crescendo of demands. Italy was on the brink of revolution.

28

The

cause for such widespread mobilization generally stemmed from the

problems related to war demobilization and conversion of war produc-

tion. Caroviveri (in£ation) and unemployment ¢gure high among the

``economic'' demands of Table 5. Among protesters, both agricultural

and industrial workers occupied a prominent position. In just two years,

between 1919 and 1920, both trade union and Socialist Party member-

ship soared. Socialist ``leagues'' in the countryside multiplied and en-

rolled an increasing number of peasants. In the 1919 national political

elections the Socialist Party succeeded in meeting the quota of 156

deputies to the Parliament, corresponding to 1,830,000 votes.

29

In Sep-

tember 1920, the General Confederation of Labor counted 1,930,000

members, more than half industrial workers.

Women performed the role of protagonists in many disputes in the data

base, both as producers and consumers. As producers, many women

had been drawn into the sphere of production during the war, to sub-

Table 5. Frequency distribution of most common demands (number of disputes)

Types of demands 1919 1920 1921 1922

Economic

Quick settlement of a dispute

Employers' unwillingness to bargain

Working conditions

Trade union rights

Contract renewal

Political demands against the state

Against layo¡s

Against violence

307

118

115

177

112

99

64

32

15

239

60

96

71

43

29

26

50

33

165

51

28

79

23

40

51

43

45

80

60

35

35

47

31

31

90

43


290

stitute for male workers called away to the front. With the end of the

war, employers either outright dismissed women to give the jobs back to

the men, or moved women to low-paid positions. In sectors such as

education, women actively struggled against their marginalization from

and in the world of production. As consumers, women were also at the

forefront of the popular struggles against in£ation (caroviveri). This type

of con£ict often turned violent, with old-fashioned forms of collective

actions typical of the ``moral economy'' such as food riots, bread riots,

assaults of bakeries, and the like.

30

While women protested to hold on to

their jobs, male war veterans and the unemployed protested for jobs.

University students, as well, protested for the decline in status that the

lack of job opportunities brought to the middle class.

Knowing the most typical actors, actions, and demands is certainly an

improvement over the rough counts of number of strikes of o¤cial strike

statistics.

31

After all, those simple counts do tell us a great deal about

what was happening in the years 1919^22.Yet, one may wonder whether

we have simply gone from one raw count (number of strikes) to many

such raw counts (e.g., number of strikes, rallies, mass meetings). For all

the richness of information, those counts are a far cry from being able to

substantiate either a view of history as agency and process,

32

or a claim

that semantic grammars allow us to collect data that address issues of

agency and process. At the very least that claim would need to show who

does what and why. For instance, what speci¢c actions did some of the

actors inTable 3 perform? What did workers do, what did employers do?

Who were the political activists? Was the increase in the frequency of

those activists related to the parallel increase in violence? To answer

those questions, let us query the database, one more time.

The data of Table 6 on the frequency distribution of disputes by types of

actions performed by workers con¢rm not only the wide repertoire of

actions workers engage in but also the rapid mobilization and demo-

bilization processes that characterized the 1919^20 and 1921^22 years.

Large-scale mobilization in the form of general and national strikes

made up much of the collective actions of the early years (91 in1919,

down to 23, 20, and 21 in 1922). Labor mobilization peaked in 1920. In

the countryside, episodes of land occupations, particularly of unculti-

vated land in the South, became more and more common during the

spring.

33

In September 1920, industrial metal workers occupied their

factories.

34

The occupation involved practically all of Italy, from such

large factories as the Fiat, Ansaldo, or Romeo works to the small semi-

artisanal workshops more typical of the Italian productive structure,

from the big cities to the small towns, from the North to the South.


291

In Lyttleton's words,

35

the factory-occupation movement was ``the

watershed between the revolutionary and reactionary phases of the post-

war crisis.'' But the ``great fear'' that traveled the country during that

month of September 1920 set o¡ a counter-mobilization process that

eventually ended with Mussolini's march on Rome on October 28, 1922,

and twenty years of Fascist dictatorship. Indeed, the ¢rst Fascist ``hit

squads'' and ``blackshirts'' started operating in the fall of 1920, right

after the factory occupation. Available evidence on the social basis of

early Fascist members shows that war veterans provided the initial back-

bone of the movement, along with college students. It was such mem-

bership that gave the movement the impression of youthfulness and

vitality.

36

It also gave the movement the know-how for violent, military-

like operations to subdue political opponents in regions where oppo-

nents were particularly strong.

37

The data of Table 7 and Table 8 on the frequency of disputes involving

Fascists clearly show a shift away from the centrality of labor relations

during the red years to the centrality of political activism during the

black years. The rise of Fascism is clearly linked to the rise of violence.

Indeed, violence was the hallmark of the early Fascist movement. Ac-

cording to Snowden,

38

Apulian Fascism was quite distinctive in relying

entirely on physical force. Distinctive, perhaps, but by no means unique.

In Tuscany, ``hit squads'' carried out both punitive expeditions and

Table 6. Frequency distribution of the most common actions performed by workers

(number of disputes)

Types of actions 1919 1920 1921 1922

Meetings

Make claims

Protest

Rally

Public speech

Strike

Occupation

Solidarity

101

83

81

4

39

78

2

11

109

37

78

11

42

160

28

19

68

7

50

5

25

60

1

12

57

0

40

2

15

11

0

7

Table 7. Frequency distribution of the actor ``Fascists'' (number of disputes)

Type of actor 1919 1920 1921 1922

Fascists 1 38 524 564


292

`èxtortion, protection rackets, and sheer plunder.''

39

In the provinces of

Ferrara

40

and Bologna

41

violence played a central role in smashing such

socialist strongholds.

42

By the 1921 elections, many local prefects in-

formed Prime Minister Giolitti that they could not guarantee order.

43

What the prefects were not saying is that much Fascist violence occurred

with police compliance if not outright collusion.

44

As Lyttleton

45

writes:

`Ìt should be remembered that the seizure of power by `force' in a

modern state is never possible, except when the army or police carries

out the coup, unless the will to resist of the Government forces has been

undermined.'' Maier

46

echoes those words when he writes that the Min-

istry of the Interiors could have mastered the forces to beat back the

Fascists, but both the police and the army were sympathetic to the

patriotic, militaristic, patriarchical, ``law-and-order'' cultural ideology

of Fascism. Furthermore, the police were the frequent target

47

of Socialist

demonstrators.

48

But, to probe further into the database, with whom did the Fascists

clash? What did the 47 occupations recorded in the data base in 1922

involve? Land, as for the peasants in 1919? Factories, as for the workers

in 1920?

The data in Table 9 provide an answer to the ¢rst question: the actors

with whom the Fascists were most likely to have violent clashes were the

Communists, in the ¢rst place, the Socialists, and the militants of the

Partito Repubblicano and Partito Popolare. Occasionally workers, peas-

ants, and strikers were the target of Fascist violence. Interestingly

enough, it appears from the data that in 1922 the police let the Fascists

run their show undisturbed, occasionally joining in with the Fascists.

49

A query on another object of Fascist violent actions (`òccupy'') shows

that contrary to the peasants' and workers' actions in1919 and 1920, the

Fascists did not occupy land or factories. They mostly occupied build-

Table 8. Frequency distribution of the actions performed by Fascists (number of disputes)

Types of actions 1919 1920 1921 1922

Clash

Wound

Kill

Assault

Destroy

Set on ¢re

Occupy

0

0

0

0

0

0

0

27

3

1

2

0

0

0

228

46

41

48

35

27

3

164

87

35

133

29

11

47


293

ings (e.g., union and socialist party locals, city halls) and, occasionally,

entire towns and cities.

50

Conclusions

In this article, I have illustrated a highly formal approach to content

analysis based on linguistics and, more speci¢cally, on semantic gram-

mars. The approach radically departs from traditional content analysis.

Typically, content analysis captures input text via a coding scheme

designed on the basis of the investigator's speci¢c substantive and theo-

retical interests. As these interests vary across investigators, so do the

content analysis schemeseven when applied to the same text. Contrary to

this traditional approach, a semantic grammar provides a tool for the

analysis of text that is invariant within a broad class of text genre,

namely narrative. In a linguistic approach to coding, the design of the

scheme does not depend upon the researcher's whims but on speci¢c

linguistic properties of the input text. Furthermore, these properties

translate into a formal notation that has a mathematical underpinning

in set theory. It is these formal characteristics of semantic grammars

that make their computer application both practical and desirable, using

widely available relational-data-base programs.

The advantages that formalism brings to content analysis, however,

come at a cost. The method achieves some of its power by narrowing

down the scope of its applicability. In fact, the method only applies to

input text organized around the basic linguistic structure noun phrase/

verb phrase typical of semantic grammars (subject, action, object ^ SAO

^ and modi¢ers). And, by and large, this means text that belongs to the

Table 9. Frequency distribution of the actors with whom Fascists ``clashed'' (number of

disputes)

Types of actors 1919 1920 1921 1922

Police

Republicans

Popolari

71

Socialists

Communists

Workers

Strikers

Peasants

0

0

0

0

0

0

0

0

2

0

0

20

2

0

1

2

13

0

7

35

145

21

0

1

1

9

9

23

72

16

4

3


294

narrative genre. This text genre is of wide interest in sociohistorical

research, particularly collective action, where many of the source docu-

ments to be analyzed are, indeed, in narrative form (e.g., police records,

newspaper articles). Nevertheless, narrative represents only a portion of

all available text genres.The application of a semantic grammar to other

types of text would be more problematic. To the extent that content

analysis has also been applied to text genres that generally do not con-

form to the simple SAO structure of narrative (e.g., in the study of

cultural indicators), the approach illustrated here is of less general

interest. Even in these cases, however, the approach should provide

researchers new insights and guidelines on how to take a research tool,

such as content analysis, away from the researcher's whims and to a

more formal and rigorous level.

In this article, I have also illustrated the power of the technique by

presenting some exploratory analyses of data that I collected from over

¢fteen thousand newspaper articles of the1919^22 period in Italy. Those

four years mark one of the most fascinating periods of Italian history.

The ¢rst two years (1919 and 1920), thered years, saw the highest level of

workers' mobilization, culminating in the widespread but unsuccessful

factory occupation movement of August 1920. The last two years (1921

and 1922), the black years, saw the successful unfolding of employers'

counter-mobilization process, culminating in a Fascist takeover of power

by Mussolini in 1922.

The available data on some 3,200 disputes, indeed, strongly show the

broad shift over time in the patterns of contention from large-scale

collective actions to small-scale political violence, from working-class

actors to right-wing goons. The data also show interesting shifts in the

matrix of actors involved in violence, from workers and the police, to

Fascists and communists. Traditional available data (e.g., strike statis-

tics) would not have provided the same richness of empirical evidence.

Traditional approaches to content analyses would not have allowed the

kind of ¢ne-grained analyses performed here. Neither would they have

allowed the queries performed on the data base in search for an answer

to a very basic question: who did what, where, when, and why?

The skeptic, of course, may wonder about the necessity of going from

words to numbers, of quantifying narratives and events, the staple of

traditional historians. Indeed, what is the point of going from words to

numbers? Answer: Because, in doing so, we move from the study of the

event ^ with its ``history of short, sharp, nervous vibrations''

51

^ to the


295

study of broader historical patterns (and the patterns revealed by the

Tables could not have been clearer). Of Braudel's three time spans of

history ^ event, conjuncture, structure ^ the method illustrated in this

article is best suited for the study of conjunctural time, ``which lays open

large sections of the past, ten, twenty, ¢fty years at a stretch ready for

examination.''

52

Indeed, the method brings out trends and similarities

from the particular. And although we certainly would not need an

apparatus of this complexity to bring out common elements among a

handful of events taken from a handful of narratives, I can hardly

imagine doing so for 3,277 events taken from 15,146 newspaper articles!

Yet a further question may lurk in the back of the skeptic's mind: is an

apparatus of such complexity really necessary? If the central methodo-

logical concern of the article is simply to obtain some information on

who did what, where, when, and why (basically the 5-W's of journalism)

do we really need to bother with mathematics and linguistics? Is this

methodological apparatus ``much ado about nothing?'' Again, the an-

swer is quite simple. For one thing, without the formal and logical

apparatus of set theory and its computer application in the form of

relational DBMS, the adoption of complex data models for the storage

and retrieval of information would be unthinkable. We would then be

back to rectangular data models, a much less £exible and much less

powerful tool.

53

More to the point, we would be back to working with

event counts, rather than event characteristics. The translation of com-

plex narrative structures into computer data models is only made possi-

ble by the adoption of the mathematical tools described here. Ultimately,

if the ``proof of the pudding is in the eating'' the empirical results

presented inTables 3 through 9 provide the best answers for the skeptics.

Certainly, o¤cial strike statistics would not have brought out such a rich

picture of those years, of the shifting patterns of actors and actions, and

of their interrelationships. And those tables are the results of very pre-

liminary, exploratory analyses of the data. They hardly do justice to the

real power of the approach. And that power lies in its ability to move

away from a concern of quantitative history with macro-level variables

(e.g., strikes, unemployment, unionization), to a concern with actors and

their actions, and the networks of their interrelationships.

54

And that is

the historian's central concern: actors and action. An approach to quan-

ti¢cation in history centered on actors and actions may go some way in

reconciling the opposite camps of qualitative and quantitative histori-

ans, between historians concerned with narrative and agency and those

concerned with models and structures.


296

Incursions into neighboring disciplines are not new in the social scien-

ces. Typically, we venture out, predatory style, quickly grabbing concepts

and tools. In his Preface to the ¢rst issue ofAnnee Sociologique, Emile

Durkheim wrote of the sociologist ``intruding as a stranger in the do-

main of the historian in order to steal, in some way, the facts that interest

him.''

55

Without proper translations of concepts from one discipline to

another, as Habermas warned us, there is a real danger that concepts

will proliferate, adding further confusion to the social science Babel of

languages. Yet, notwithstanding these dangers, there are also clear ad-

vantages in moving into the territory of other disciplines. Consider the

concept of semantic grammar, reappropriated here from linguistics. The

advantage of venturing into linguistics is hardly con¢ned to the use of

the concept of ``semantic grammar.'' Perhaps an even greater advantage

comes from the wide variety of tools that one can ¢nd in that ¢eld.

Indeed, venturing into the territory of the linguist opens a door to a

great deal more than semantic grammars. For instance, linguists will tell

us that the question of meaning and text understanding is no simple

issue, even for very simple prose. That in itself is su¤cient to debunk the

recommendation of traditional content analysis that all coding rules

should be clearly spelled out in a handbook of coding instructions.

Laudable recommendation ^ having clear and uniform rules is certainly

better than having no rules; but it misses the point. No handbook,

however detailed, could ever come close to providing the rules that go

into understanding text. In linguistics one will also ¢nd, ready-made,

theories of text genres and of narrative. The application of semantic

grammars to narrative texts can thus be put into wider contexts. Finally,

the linguistic theory of semantic roles will further tell us that the SAO

structure adopted here has limitations even within narrative.

56

Quick

incursions into neighboring disciplines will ¢ll up our robbers' bags in

the short term, but long-term settlements are likely to enrich the debate

much more permanently and much more solidly.

All of this may still leave the ``theoretical skeptics'' shaking their heads.

After all, theory is nowhere to be seen in this article; not the mathemat-

ical and linguistic theories upon which the methodological work pre-

sented here rests. We have seen plenty of that here (what is method in

one ¢eld is theory in another). But what about sociological theory? The

point is well taken. No doubt, the substantive and theoretical underpin-

nings of the methodological developments of this article are not clearly

spelled out.Yet, the reader should not be deceived by a primary concern

with methodology and by a focus on exploratory and descriptive data

analyses. The article, in fact, tackles two broad theoretical problems.


297

Within history, the problem of interpretation of Italian Fascism has long

haunted historians and it is still hotly debated. For instance,was Fascism

a response to a generalized malaise generated by a war that cost Italy

over 600,000 soldiers, or was Fascism a reaction to Socialism, i.e., a

reaction to the widespread working-class mobilization? In which case,

speci¢cally, reaction by whom and to whom, as De Felice

57

aptly put it:

land owners to radicalized agricultural laborers? industrial entrepre-

neurs to the continuous threat of strikes, in particular, the August 1920

factory-occupation movement? members of the petty bourgeoisie to the

steady erosion of their status and to the decline of labor-market oppor-

tunities? The answer to these questions is likely to stem from a careful,

empirical investigation of the interests and actions of speci¢c classes and

class fractions, indeed, who did what, where, when, and why. A view of

history as agency and process and the inadequacy of both traditional

data and methods to substantiate that view originally pushed me to

explore alternative methodological approaches in the quantitative study

of history. My work on Italian strikes in the period 1950^1980 based on

o¤cial strike statistics and econometric models taught me a simple

lesson: to understand what workers do (e.g., strikes), at the very least,

we also need to understand what employers and the state do. The

methodological developments illustrated here are the results of those

substantive concerns.

58

At the cross-roads of social sciences and history, no less elusive is the

problem of a general interpretation of Fascism, and not just of Italian

Fascism. Is Fascism one of the three paths to the modern world, as

Barrington Moore claimed,

59

or is it just a temporary and unstable

solution to the problems of modernization as Rueschemeyer, Stephens,

and Stephens

60

would have it? In either case,what are the social bases of

Fascism? Which speci¢c class alliances make Fascism possible? That

among the landed upper class, the state and the bourgeoisie, as in

Barrington Moore's view,

61

or that between the industrial elite and the

petty bourgeoisie, as in someMarxist interpretations?

62

At a yet more general level, do theories of mobilization and counter-

mobilization have anything to o¡er for the understanding of the years

between 1919 and 1922?

63

Can these theories account for the widespread

working-class mobilization of the red years and the Fascist reaction of

the black years? What explains the ultimate failure of the one and the

success of the other? What resources (material, cultural, political) could

the two movements muster? What opportunities did the political envi-

ronment (domestic and international) o¡er to various social groups?,


298

How did the organizations involved (in particular, the Socialist Party

and the Fascist Party) frame the issues so as to sell them successfully to

their constituencies?

64

No doubt, the stormy period of criss-crossing mobilization of various

groups (students, war veterans, women, agricultural laborers, industrial

workers, white-collar workers) during1919^20 and counter-mobilization

during 1921^22 has the characteristics of Tarrow's ``cycles of protest.''

65

After all, these cycles are characterized by the rapid di¡usion of collec-

tive action from more to less mobilized groups, by the great innovation

in the forms of collective action, by the combination of organized and

unorganized protest, and by the intensi¢ed interaction between chal-

lengers and authorities. During cycles of protest di¡erent groups join

the mobilization process at di¡erent points. Early risers trigger a variety

of processes of di¡usion, imitation, and reaction among other groups.

66

The evidence from the 1919^22 period in Italy goes hand in hand with

the evidence on another such cycle of protest, the 1968^72 strike wave in

Italy.

67

The comparative study of the two cycles is likely to yield impor-

tant theoretical insights.

These are the kinds of historical and sociological questions that animate

this project in the long-term. The empirical evidence for that view and

for those questions is unlikely to come solely from the matrix of actors

and actions that the method presented here can provide. Nodoubt, more

is needed.

68

Alas, there is no catch-all method in the study of complex

historical questions. But, undoubtedly, the method presented here repre-

sents a step forward toward the historians' view of history as agency. In

Elton's words: ``Despite attempts to deny this, it [quantitative history]

can e¡ectively operate only by suppressing the individual ^ by reducing

its subject matter to a collectivity of human data in which the facts of

humanity have real di¤culty in surviving.''

69

In summary, despite the complexity of the formal representation of a

semantic grammar (subjects are related to actions, actions to objects,

subject modi¢ers to subjects, action modi¢ers to actions, and sets of

subjects, actions, and objects are hierarchically linked into higher-level

aggregates, such as events and disputes), from a methodological stand-

point, the approach presented here has several advantages over tradi-

tional content analysis. First, the linguistic approach is highly formal-

ized. It escapes the ad-hoc nature of much traditional content-analysis

research. Second, it provides re-usable data, as the data are coded not on

the basis of the investigator's interests but on the basis of linguistic


299

properties intrinsic to the text.Third, it provides much richer data. Fourth

the properties of the method are desirable not only in the abstract. The

method, in fact, is highly practical. It can easily be implemented within

existing and widely-di¡used relational-data-base management systems.

The exploratory analyses of data gathered from newspapers on collec-

tive action in Italy during the 1919^22 period con¢rm the power of the

technique for use in large-scale sociohistorical research. Finally, perhaps

the best way to gauge the worth of a method is in comparison to

available alternatives. On that score, the approach to the collection and

analysis of event data described here provides information that is either

simply unavailable in other types of data (e.g., strike statistics) or is

structured in far more sophisticated ways (namely, in relational form).

These characteristics, in turn, allow investigators to test hitherto untested

or untestable theoretical propositions, or to bring empirical evidence to

bear on novel or unsupported substantive issues. Substantive and theo-

retical innovation should be the ultimate touchstone of good method-

ology.

Acknowledgments

The methodological and substantive project described in this article has

been supported by: Con¢ndustria, Consiglio Nazionale delle Ricerche

(progetto strategico ``Il con£itto e le relazioni di lavoro nel prossimo

decennio''), Formez, Intersind, the National Science Foundation (grants

SES^8511632 and SBR^9411739), and the University of Wisconsin-Mad-

ison. I would like to thank Simona Colarizi and the Editors ofTheory

and Society for their help with the substantive aspects of this article.

Notes

1. R.W. Fogel, ``The Limits of QuantitativeMethods in History,''The AmericanHistor-

ical Review 80 (1975): 329^350, 345.

2. P. Spriano,L'occupazione delle fabbriche (Turin: Einaudi, 1964).

3. G. Maione,Biennio Rosso, Autonomia e spontaneita© operaia nel 1910^1920 (Bologna:

Il Mulino, 1975).

4. The word narrative, of course, conjures up images in the minds of historians of

opposed and entrenched scholarly camps, of academic divisions between those who

view narrative as the proper way to study history and those who believe that the only

way to turn history into a science is by abandoning narrative (after all, a tool of

storytelling) and adopting instead formal mathematical/statistical models (see the

signi¢cant title of Fogel and Elton's book,Which Road to the Past? Two Views of

History, New Haven: Yale University Press, 1983). That I should propose to use


300

narrative for quantitative purposes (``to go from words to numbers,'' as I have put it)

may end making no one happy or, worse yet, may give the quantitative scholars the

feeling of victory, as they manage to turn their opponents' main tool into their own

tool; far from me the idea of a ``scienti¢c'' approach to history devoid of narrative

and rhetoric and that of narrative as the exclusive world of storytelling (I take up

these issues in two book manuscripts I am currently working on: R. Franzosi,A

Trilogy of Rhetoric: The Rhetorical Foundations of Social Science QuantitativeWork,

Cambridge: Cambridge University Press, forthcoming; R. Franzosi,From Words to

Numbers: Narrative as Data, Cambridge: Cambridge University Press, forthcoming;

R. Franzosi, ``Narrative Analysis,'' in J. Hagan, editor,Annual Review of Sociology

23, forthcoming.

5. For a review, see R. Franzosi, ``The Press as a Source of Socio-Historical Data:

Issues in theMethodology of Data Collection fromNewspapers,''Historical Methods

20 (1987): 5 16; for another recent example, see S. Tarrow,Democracy and Disorder.

Protest and Politics in Italy, 1965^1975 (Oxford: Clarendon Press, 1989).

6. For example, see C. Tilly,The Contentious French (Cambridge: Harvard University

Press, 1986); C. Tilly, Popular Contention in Great Britain, 1758^1834 (Cambridge:

Harvard University Press, 1995).

7. I conceived the idea of using newspapers as sources of data for the study of the1919^

22 period as a postdoctoral fellow at the Center for Research on Social Organization

of the University of Michigan directed by Charles Tilly. It is the way Tilly organized

data collection from newspapers for his Great Britain Study in terms of formations

and actor phases (i.e., actors and actions) that eventually led me to discover semantic

grammars as tools of data collection.

8. V. Propp,Morphology of the Folktale (Austin: University of Texas Press, 1968 [1928]).

9. A. Greimas,Semantique Structurale. Recherche de Methode(Paris: Larousse, 1966).

10. W. Labov and J. Waletzky, ``Narrative Analysis,'' in J. Helm, editor,Essays on the

Verbal and Visual Arts (Seattle: University of Washington Press, 1967), 12^44; see

alsoW. Labov,Language in the Inner City (Philadelphia: University of Pennsylvania

Press, 1972).

11. M.A.K. Halliday, ``Language Structure and Language Function,'' in J. Lyons, editor,

New Horizons in Linguistics (Harmonsworth: Penguin Books Ltd., 1970), 140^165;

M.A.K. Halliday,Language as Social Semiotic (London: Arnold, 1978).

12. D. Rumelhart, ``Notes on a Schema for Stories,'' in D. Bobrow and A. Collins, editors,

Representation and Understanding (NewYork: Academic Press, 1975), 211^236.

13. J. Mandler, ``A Code in the Node: The Use of Story Schema in Retrieval,''Discourse

Process 1 (1978): 14^35.

14. T. van Dijk, ``Structures of News in the Press,'' in T. van Dijk, editor,Discourse and

Communication. New Approaches to the Analyses of Mass Media, Discourse and

Communication (Berlin:Walter de Gruyter, 1985), 69^93, 85.

15. Labov andWaletzky, ``Narrative Analysis.''

16. Halliday, ``Language Structure and Language Function.''

17. Initial, terminal, and non-terminal symbols provide, respectively, the starting symbol

of a rewrite chain, the last symbol of that chain, i.e., a symbol that cannot be further

rewritten and symbols that can be further rewritten via the application of rewrite

rules.

18. On the factory occupation movement, see Spriano,L'occupazione delle fabbriche.

19. A. Lyttleton, ``Revolution and Counter-Revolution in Italy, 1918^1922,'' in C. Bertrand,

editor, Revolutionary Situations in Europe, 1917^1922: Germany, Italy, Austria-Hun-

gary (Montreal: Interuniversity Centre for European Studies, 1977), 63.


301

20. For an introduction to the general problems of the rise of Fascism in Italy, see R. De

Felice,Mussolini il Fascista,Vol. 1. La conquista del potere, 1921^1925(Turin: Einaudi,

1995 [1966]; R. DeFelice,Le interpretazioni del Fascismo (Bari: Laterza, 1989 [1969]);

A. Lyttleton,The Seizure of Power: Fascism in Italy. 1919^1929 (New York: Charles

Scriber and Sons, 1973); C. Maier,Recasting Bourgeois Europe (Princeton: Princeton

University Press, 1975). For an analysis of speci¢c, local situations, in the English

language, see P. Corner, Fascism in Ferrara 1915^1925 (Oxford: Oxford University

Press, 1975); A. Kelikian,Town and Country Under Fascism: the Transformation of

Brescia, 1915^1926 (Oxford: Oxford University Press, 1986); F. Snowden,Violence and

the Great Estates in the South of Italy: Apulia, 1900^1922 (Cambridge: Cambridge

University Press, 1986); F. Snowden,The Fascist Revolution in Tuscany, 1919^1922

(Cambridge: Cambridge University Press, 1986); D. Bell,Sesto San Giovanni:Workers,

Culture and Politics in an ItalianTown, 1880^1922(New Brunswick: Rutgers University

Press, 1986); A. Cardoza,Agrarian Elites and Italian Fascism:The Province of Bologna

1901^1926 (Princeton: PrincetonUniversity Press, 1982).

21. R. Franzosi, ``FromWords to Numbers: A Set Theory Framework for the Collection,

Organization, and Analysis of Narrative Data,'' in P. Marsden, editor,Sociological

Methodology 24 (Oxford: Basil Blackwell, 1994), 105 136.

22. See for all, R. Franzosi, ``FromWords to Numbers: A Generalized and Linguistics-

Based Coding Procedure for Collecting Event-Data from Newspapers,'' in: C. Clogg,

editor, Sociological Methodology, 19 (Oxford: Basil Blackwell, 1989), 263^298; Fran-

zosi, ``From Words to Numbers: A Set Theory Framework for the Collection,

Organization, and Analysis of Narrative Data.''

23. R. Franzosi, ``A Sociologist Meets History. Critical Re£ections upon Practice,''The

Journal of Historical Sociology 9 (1996): 354^392.

24. For a good introduction to narrative from a linguistic perspective, see G. Genette,

Narrative Discourse: an Essay in Method (Ithaca: Cornell University Press, 1980); G.

Genette, Narrative Discourse Revisited (Ithaca: Cornell University Press, 1988); S.

Cohan and L. Shires,Telling Stories: aTheoretical Analysis of Narrative Fiction(New

York: Routledge, 1988); S. Rimmon-Kenan,Narrative Fiction: Contemporary Poetics

(London: Methuen, 1983).

25. For a re£ection upon the development of my research agenda, see Franzosi, ``A

Sociologist Meets History.''

26. See R. Franzosi,The Puzzle of Strikes: Class and State Strategies in Postwar Italy

(Cambridge: Cambridge University Press, 1995).

27. E. Le Roy Ladurie,The Territory of the Historian (Chicago: University of Chicago

Press, 1979), 3.

28. Lyttleton, ``Revolution and Counter-Revolution in Italy, 1918 1922,'' 63.

29. Spriano,L'occupazione delle fabbriche, 16^18.

30. For example, see Kelikian,Town and Country Under Fascism: the Transformation of

Brescia, 1915^1926, 82; Snowden,Violence and the Great Estates in the South of Italy:

Pulia, 1900^1922, 164.

31. On the limitations of these data, see M. Shalev's classical piece, ``Lies, Damned Lies

and Strike Statistics: The Measurement of Trends in Industrial Con£icts,'' in C.

Crouch and A. Pizzorno, editors,The Resurgence of Class Con£ict inWestern Europe

Since 1968,Volume 1, National Studies (NewYork: Holmes & Meier, 1978), 1^20.

32. See Franzosi,The Puzzle of Strikes.

33. Snowden,Violence and the Great Estates in the South of Italy: Apulia, 1900^1922.

34. Spriano, L'occupazione delle fabbriche; Maione, Biennio Rosso, Autonomia e sponta-

neita© operaia nel 1910^1920, 179^190.


302

35. Lyttleton,The Seizure of Power: Fascism in Italy, 1919^1929, 36.

36. Lyttleton, ``Revolution and Counter-revolution in Italy, 1918 1922,'' 71^72; Snowden,

Violence and the Great Estates in the South of Italy: Pulia, 1900^1922, 183 185;

Snowden,The Fascist Revolution inTuscany, 1919^1922, 157^160.

37. For an example, see Snowden,The Fascist Revolution inTuscany, 1919^1922, 80.

38. Snowden,Violence and the Great Estates in the South of Italy: Pulia, 1900^1922, 186^187.

39. Snowden,Violence and the Great Estates in the South of Italy: Pulia, 1900^1922, 80, 158^

160.

40. Corner,Fascism in Ferrara 1915^1925, 138^143.

41. Cardoza, Agrarian Elites and Italian Fascism: The Province of Bologna 1901^1926,

308, 314, 346^348.

42. On the role of violence in early Fascism, see De Felice,Mussolini il Fascista,Vol. 1, La

conquista del potere, 1921^1925, 3^99.

43. Maier,Recasting Bourgeois Europe, 315.

44. Maier,Recasting Bourgeois Europe, 177; Corner,Fascism in Ferrara 1915^1925, 119 120;

Kelikian, Town and Country Under Fascism: the Transformation of Brescia, 1915^

1926, 142 143; Snowden,The Fascist Revolution inTuscany, 1919^1922, 198.

45. Lyttleton,The Seizure of Power: Fascism in Italy, 1919^1929, 86.

46. Maier,Recasting Bourgeois Europe, 321.

47. Maier, Recasting Bourgeois Europe, 177, 317; Snowden, The Fascist Revolution in

Tuscany, 1919^1922, 195.

48. Agrarian landlords, threatened in their material interests by the high wage increases

and high militancy of agricultural laborers and by eroding rents in the face of long-

term contracts and high in£ation, were instrumental in providing much of the

support (e.g., trucks, money) necessary for the ``punitive expeditions'' of urban thugs.

It may well be that the agrarian elites saved ``Mussolini's urban Fascism from

political extinction'' (Corner,Fascism in Ferrara 1915^1925, x; Lyttleton,The Seizure

of Power: Fascism in Italy, 1919^1929, 53^55). As Cardoza argues, Fascism in the

Bologna province remained urban and limited to anti-socialism, with little elite

support. In Spring 1921, agrarian o¡ensives made it a mass movement (Agrarian

Elites and Italian Fascism: The Province of Bologna, 1901^1926), 314. The urban

bourgeoisie had been more reluctant (at least until the factory occupation) to em-

brace the Fascist ideology and its tactics. Despite the urban origin of Fascism

(Lyttleton,The Seizure of Power: Fascism in Italy, 1919^1929, 51). Fascist inroads into

the urban world of industry came late and slow. At the polls, Fascism failed to win

over more than a small fraction of the urban population; seeW. Brustein, ``The Red

Menace and the Rise of Italian Fascism,''American Sociological Review 56 (1991):

652^664, 653. The provinces in which Fascism had the greatest electoral success were

all primarily agricultural. Kelikian, in his study of the local situation in Brescia,

shows that agrarian terrorism in the Brescian countryside started as early as the

winter of 1920 with ``punitive expeditions'' by imported ``blackshirts'' (Kelikian,

Town and Country Under Fascism: the Transformation of Brescia, 1915^1926, 133,

142 143, 151^153). But it is not until 1923 that Fascism started making incursions into

Brescian industry (Kelikian,Town and Country Under Fascism: theTransformation of

Brescia, 1915^1926, 144, 176). Similarly, Bell shows that in the industrial town of Sesto

SanGiovanni, one of Italian labor's traditional strongholds, the Fascists relied less on

the use of violent tactics than on cultural forms (Bell,Sesto San Giovanni:Workers,

Culture and Politics in an Italian Town, 1880^1992, 160). As late as April 1921, the

Sesto fascio apologized for a Milanese Fascist attack (Bell, Ibidem,162).The ¢rst raid

against the Socialist Party headquarters did not occur until September 1922, a month


303

before the ``march on Rome.'' Only on October 31, after the ``march on Rome'' did

Fascists from Sesto, Milan, and the Lomellina occupy the Sesto town hall (Bell,

Ibidem, 177^178.)

49. For evidence, see also Corner, Fascism in Ferrara 1915^1925, 119^120, 201 202;

Cardoza, Agrarian Elites and Italian Fascism: The Province of Bologna 1901^1926,

348; Kelikian, Town and Country Under Fascism: the Transformation of Brescia,

1915^1926, 142 143; Snowden,The Fascist Revolution inTuscany, 1919^1922, 198.

50. For similar evidence, see Lyttleton,The Seizure of Power: Fascism in Italy, 1919^1929,

80; Maier,Recasting Bourgeois Europe, 335^337; Kelikian,Town and Country Under

Fascism: theTransformation of Brescia, 1915^1926, 157^158.

51. F. Braudel,On History (Chicago: The University of Chicago Press, 1980), 3.

52. Braudel, On History, 27; see also Le Roy Ladurie,The Territory of the Historian,

111^132.

53. Franzosi, ``From Words to Numbers: A Set Theory Framework for the Collection,

Organization, and Analysis of Narrative Data.''

54. For an example, see the work byTilly in this issue.

55. E. Durkheim, ``Preface,''Annee Sociologique 1 (1898): i^iii, iii.

56. On these issues, see Franzosi,FromWords to Numbers: Narrative as Data; Franzosi,

``Narrative Analysis.''

57. De Felice,Mussolini il Fascista, Vol. 1,La conquista del potere, 1921^1925, 174.

58. For a re£ection upon my research agenda, see Franzosi, ``A sociologist meets history.''

59. B. Moore, Social Origins of Dictatorship and Democracy (Boston: Beacon Press,

1966), 433^452.

60. D. Rueschemeyer, E.H. Stephens, and J.D. Stephens,Capitalist Development and

Democracy, (Chicago: University of Chicago Press, 1992), 152.

61. Moore,Social Origins of Dictatorship and Democracy, 436.

62. For example, see N. Poulantzas, Fascism and Dictatorship (London: Verso, 1979

[1970], 85^88, 237^246.

63. For a good introduction to various theories of social movements, see D. McAdam, J.

McCarthy, and M. Zald, editors, Comparative Perspectives on Social Movements

(Cambridge: Cambridge University Press, 1996).

64. On the ability of the early Fascist Party drastically to change their strategy to appeal

to di¡erent constituencies in the aftermath of the 1920 electoral defeat, see E.Gentile,

Storia del partito Fascista, 1919^1922, Movimento e milizia(Bari: Laterza, 1989), 60^162.

65. S.Tarrow,Power in Movement (Cambridge: Cambridge University Press, 1994), 153 169.

66. Tarrow,Power in Movement, 155 156.

67. See Franzosi,The Puzzle of Strikes, 257^342; see also Tarrow,Democracy and Dis-

order. Protest and Politics in Italy, 1965^1975.

68. For example, the results of the political and administrative elections of1919, 1920,

1921, and 1922 as in A. Szymanski, ``Fascism, Industrialism, and Socialism: The Case

of Italy,''Comparative Studies in Society and History15 (1973): 395^404, or Brustein,

``The Red Menace and the Rise of Italian Fascism;'' not to mention a more qualita-

tive investigation of the cultural and ideological frames of various organizations.

69. G.R. Elton, ``Two Kinds of History'' in R.W. Fogel and G.R. Elton,Which Road

to the Past? Two Views of History (New Haven: Yale University Press, 1983), 71^122,

118^119.

KAP ^ RYSO 399 ^ PIPS Nr. 125454 Rapati 9-5-97 09:48 4

304

mobilization and counter-mobilization processes: from the “red years” (1919–20) to the...

Documents