Locating the Cognitive Bottleneck in Novice Function Comprehension: A Literature Synthesis and Diagnostic Instrument Design Across Python, C/C++, and Scratch My Blocks

Function introduction is one of the most frequently identified bottlenecks in introductory programming, yet the literature does not converge on which sub-aspect of function is the load-bearing one for novices. Four bottleneck candidates dominate the discussion: (1) parameter passing semantics, (2) return value handling, (3) scope and lexical visibility, and (4) abstraction intent — the why of decomposing a procedure into a named reusable unit. A fifth axis, recursion, is co-introduced in some curricula and interacts with the first four in characteristic ways. This paper synthesises four decades of cognitive-science and computing-education literature to produce three contributions. First, a verdict on the load-bearing candidate: abstraction intent (candidate 4) is the dominant under-studied bottleneck at the THCS level, despite the field's disproportionate emphasis on parameter and scope mechanics; the candidate that visibly fails on assessments is parameter passing or return, but the candidate that causally governs durable function competence is intent. Second, a language-family analysis showing that the load-bearing candidate shifts: in Python the return-byobject-reference model loads parameter passing; in C/C++ the declaration-vs-definition split and pointer/reference duality loads parameter passing differently; in Scratch My Blocks the absence of return values redirects the entire load onto intent and sideeffect modelling. Third, a diagnostic instrument design — a think-aloud, worked-example, and transfer-error battery — with 22 exemplar diagnostic items across the five candidates, each grounded in a published learner error pattern, together with a scoring rubric and a validity argument that anchors face validity to published exemplar errors rather than fresh fieldwork. Per the elix-researches no-cohort scope, the instrument is described and theoretically validated; deployment is identified as a downstream cohort node (B1.4 Python-first cohort, A3.1 variable mental model, C2.1 PRIMM cohort) and is out of scope for this paper.