More information on the applications of WORDS may be found here:

Cary, Charles D. \"Software Review: A User's View of WORDS.\" Computers and the Humanities 13.2 (1979): 129-30. Web.

Iker, Howard P. and Robert H. Klein. \"WORDS: A Computer System for the Analysis of Content.\" Behavior Research Methods & Instrumentation 6.4 (1974): 430-438. Web.

Program Overview

WORDS was a system of content analysis programs developed by Howard P. Iker and associates at the University of Rochester (Cary 129). It was designed to identify major themes within a text without relying on preexisting categorization systems (Iker and Klein 430).

The WORDS system was first described in the mid-1960s and was in continuous development over the 1960s and 1970s (Iker and Klein 430). The name WORDS is a mnemonic referring to the collection of programs within the system and \"is based upon the pivotal logic of the method itself, viz, that sufficient information exists within the word and within the temporal associations among and between words to allow data-generated elicitation of major content themes and materials\" (430). As of 1979, WORDS had been installed in over a dozen facilities across North America and continental Europe, in addition to South America (Cary 130).

As a system, WORDS was composed of a collection of compatible and interrelated but otherwise independent programs designed to help users to manage the data flow of their analysis by both passing and receiving data as needed (Iker and Klein 430). It contained a suite of 38 sub-programs, 32 which were available to be called by users, with the remaining 8 designed to be called only by other programs or act as service programs (Cary 129).

Each program within WORDS intended for human access was assigned a mnemonic used to call it, such as CLOSE to output data and retrieve any messages generated, or JOIN to concatenate data sets (Iker and Klein 432). Users could set a run of programs and call them in any order (432). WORDS included an associated program, JOBS, which produced diagnostic error messages and assisted users in communicating with the computer's operating system, which removed the need for users to learn how to do so themselves (432). JOBS also included five fixed configurations of WORDS programs, to save time for users when calling programs that were usually preceded or followed by particular other programs (432). Iker and Klein described JOBS as

our best current attempt to make the use of WORDS a relatively straightforward task requiring a minimum of computer sophistication. It is further developed to minimize the problems of implementing WORDS at installations other than that at which the system was developed. Our experience has been that JOBS can substantially reduce the effect of installation differences through use of its many options and parameters. (432)

WORDS used individual words as units of information and divided the text according to units of time, segments of a specified length, paragraphs or other similar divisions (Iker and Klein 430). Within a text's divisions, WORDS determined word frequencies and frequencies of word correlations to develop an intercorrelation matrix and systematically locate common word groups (430). It was designed specifically to avoid dependency on predetermined categories or dictionaries, and instead to generate its own thematic \"centroids\" (430). According to Iker and his colleague Robert H. Klein, this design approach was a deliberate response to research findings:

Our research has demonstrated that these word groups both represent and indentify the major content themes in the raw data. Elaboration and analysis of these thematic data can then be conducted by using such methods as cluster of factor scoring to locate and inspect highly saturated sections of the raw data, by configurational analyses to examine changes in content profile patternings in different portions of the data, and so on. (430)

WORDS worked by manipulating words statistically and handled words individually while also associating metadata about that word such its location, the part of speech it represents, or the speaker (Iker and Klein 431). It separated words and punctuation marks into an independent standardized record with 10 pieces of associated metadata (431).

Users were required to prepare the text before processing (431). Preparation entailed selecting an observational unit, such as paragraphs, or segments split by time such as one minute's worth data (431). Iker and Klein found units of one minute particularly useful for psychotherapy data (431), though their 1974 paper does not indicate whether units of time refer to how much the computer can process in that span, or an equivalent quantity of spoken language, or a specific number of sentences or words. Users then needed to assign a unique number to each unit and transfer the data onto punched cards (431). For each card, contractions needed to be expanded into separate words, proper nouns flagged by preceding them with '$', and quotation marks either removed or replaced with double apostrophes (431). All text on a given card had to belong to the same segment, though the segment could span as many cards as needed (431).

Iker and his colleagues developed WORDS in PL/I, and as of the 1974 edition it would run on

an IBM S/360 Model 50 (or higher) computer with at least 256K bytes of available core, which operates under full OS, has a PLII compiler of LevelS or greater, and has available a 2314 disk device or interchangeable counterpart. If these requirements can be satisfied, WORDS can be implemented at the target installation. The project has several memoranda which detail specific requirements; they are available on request. (Iker and Klein 437).

Charles D. Cary, who reviewed WORDS for Computers in the Humanities in 1979, found that WORDS required a specialist to install it and assist with any problems encountered in its use (Cary 130). Cary recommended that users of the WORDS system be familiar with operating packaged programs, and have a basic understanding of statistics, as this experience would \"help to make sense out of what is happening during the content analysis and to interpret the results\" (130).

Indeed, WORDS required a large investment of time despite efforts to make it as simple as possible (Iker and Klein 437). While not \"intolerably difficult\", Iker and Klein

wish[ed] to stress that one should not expect to make use of the system as though it could be acquired in one day, run up on the target computer on the next, and put into operation on the third. Understanding the interrelationship among and between programs, learning to use the JOBS program, making changes to satisfy installation differences by appropriate option and parameter changes, all require substantial amounts of time. While the project itself sets no constraints on potential users, we have always advised interested researchers to consider carefully whether they wish to invest the time to learn to use WORDS effectively. (437)

Iker and Klein emphasized the value of WORDS as a research tool despite the challenges of using it. In their view,

[t]he asset of WORDS is that the themes and content areas it produces are much richer, more detailed, and, by definition, more in tune with the original data than those which would be produced by the sole use of a priori category systems. At the same time, one of the liabilities of WORDS is that these results are inevitably defined in the words of the speaker. This is a reasonable characteristic so long as WORDS is used to describe solely the data base under analysis. When, however, one tries to generalize from one set of data to another, to compare one data base with another, even to compare two separate analyses on different parts of the same data, substantial problems can arise, because the results are difficult to put onto a common measuring dimension or set of dimensions. (Iker and Klein 436)

From the user's perspective, Cary found the manual detailed, with over 200 pages of material. He noted its seven-point write-up for each user-called program, the full listing of the diagnostic messages produced in response to system errors, a glossary of data processing terms as used in the manual, and example configurations of the user-called programs for content analysis (Cary 130). The manual also included installation memoranda and listed a variety of texts describing the system and research applying it (Cary 129). Though WORDS was designed for use with English text, Cary found that researchers had applied it to Russian, French and a South African dialect, Chichewa and that it could be applied to additional languages as long as the text used the Latin alphabet and Arabic numerals (Cary 130).

Sources

Cary, Charles D. \"Software Review: A User's View of WORDS.\" Computers and the Humanities 13.2 (1979): 129-30. Web.

Iker, Howard P. and Robert H. Klein. \"WORDS: A Computer System for the Analysis of Content.\" Behavior Research Methods & Instrumentation 6.4 (1974): 430-438. Web.