Formulaic sequences (FSs), or prefabricated multi‐word structures are frequently difficult to distinguish objectively, and actual corpus‐driven methods produce structurally incomplete, overlapping, or overly expanded structures of questionable psychological validity and pedagogical usefulness. To handle these restrictions, this inquiry assessed transitional probability as a potential metric to ameliorate the identification of FSs by proposing 100 four‐word sequences from the British National Corpus, differing in transitional probabilities between words, to native and non‐native speakers of English (N = 293) in a sequence completion task. Results disclosed that the application of transitional probability decreases many of the problems related to current approaches to FS identification and can generate lists of FSs that are more functionally salient and psychologically valid.