Content analysis in general, and computerized content analysis in particular, have often been criticized for making heroic assumptions. First, whether one measures the frequency with which particular words or phrases appear in a given text or set of texts (as in the Yoshikoder tool described below), or whether one measures the centrality of words or phrases in a rhetorical network (as in the Crawdad tool described below), one presumes that their meaning is stable over time – or at least sufficiently stable to permit analysis – and can be grasped without sensitivity to context. Second, because all content analysis tools necessarily analyze what has been said or printed, they assume that what is articulated is of greater substantive import than what is not articulated, and they do not help the analyst probe the meaning of silence.

These critiques are fair, but content analysis still has utility, and it can be productively joined to methods like discourse analysis. Concerns about the instability of meaning can be alleviated by analyzing a narrower range of texts – from a single or relatively homogeneous set of speakers, from a relatively short time span, from a single country – and by generating search-terms and coding rules based on context-sensitive reading of select texts and secondary literature. Analysts sensitive to the silences within texts might employ content-analytic methods to explore explicit articulations and then turn to more critical methods to identify and probe crucial silences. Content analysis is a blunt but suggestive method, and it necessarily leaves it to the analyst to interpret the findings. But content analysis can reinforce, or potentially challenge, the qualitative analyst's impressionistic observations about what linguistic patterns are more or less common. It can thereby strengthen discourse analyses, which normally rest on assertions that a particular discourse, or set of elements, is commonplace or even dominant (“the main signifying elements of the discourse … must be identified”), as are certain “chains of connotations among these signifying elements.” Content analysis, whether conducted by computers or human beings, is never sufficient on its own, but it has an important role to play in the study of language – as long as analysts do not rest conclusions exclusively on it, as long as analysts combine it with other methods and approaches, and as long as analysts are aware of its limits.