Is memory really comprised of separate structures or types (semantic v episodic, ltm v stm, phonological loop v visuospatial sketchpad, etc), or just one kind of memory that functions in different ways when conditions (e.g., type of stimulus, length of time) vary?
The levels (or depth) or processing (LOP) perspective (Craik & Lockhart, 1972; Craik & Tulving, 1975) is a unistore account that replaces the classic STM/LTM distinction with consideration of whether the processing of to-be-remembered information is shallow (i.e., processing physical or phonological features) versus deep (or processing of meaning). Information is remembered better and longer under conditions of deep elaborative rehearsal (i.e., linked to existing knowledge) than when processed only with shallow maintenance rehearsal (i.e., phonological or articulatory repetition).
LOP was attacked as circular and descriptive rather than explanatory, and was generally dismissed as an alternative to multistore models. Nevertheless, its empirical phenomena are robust and important. For recent work, see: Rose, Myerson, Roedigger & Hale (2010): "Similarities and differences between working memory and long-term memory: Evidence from the levels-of-processing span task." Journal of Experimental Psychology: Learning, Memory, and Cognition 36.2 (2010): 471-483. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2832224/