The importance of CAH is noted by Romero and Manjarres (2017):
Within CAH there are two positions: the strong (a priori) and the weak (posteriori) versions.
The strong version holds that predictions about a learner’s ability to learn language features can be constructed by comparing the two languages.
The strong version of CAH is considered, according to Ronald Wardhaugh (1970), to be “unrealistic and impracticable” (p. 124).
The weak version analyzes the recurring errors that a learner makes in their output due to interference (Wardhaugh, 1970, p. 123).
Eventually, the weak version gained credence over the strong view and was later incorporated into the development of EA (Gass, 2013, p. 87).
Early days:
"Inferred in their studies that there was a little advantage of administering error analysis which exploited the concept of negative transfer in second language learning. They held that native language interference constituted nothing but a tiny portion which caused errors in the output. Therefore, CA was regarded as being irrelevant to second language acquisition in North America at that time" (p. 100).
At the root of the criticism towards CAH are two concepts: the difficulty of learning, and its claim to predict errors (Gass, 2013, p. 87). Cracks in CAH began to appear when researchers began to apply CAH and noticed that many of the errors that were being made were not being caused by negative transfer from the L1.
While it is unclear what causes these errors to occur, Jimenez (2005, p. 243) suggests it is generally considered to be intralingual errors. In addition to failing to explain why errors occurred, CAH’s notion that elements of language that were identified as being different and therefore difficult also disintegrated. Researchers, such as Chan (2004), found that many of the areas that CAH identified as being difficult were not accurate. Instead, there were language features that were considered ‘difficult’ according to CAH that the learners acquired easily.
While past work has been done to discredit CAH, more recent work conducted by individuals such as Kellerman and Smith, and Oldin and James (as cited by Chan, 2004, p. 56) have updated and revitalized the field. In addition to CAH, Error Analysis (EA) is another type of linguistic analysis that began to be developed in the 1970s in response to the failures and criticisms of the CAH.