Clinical recognition memory testing in older adults should move beyond accuracy alone

In 1972, Endel Tulving proposed a unique type of memory that encapsulated the human experience 9 of remembering the "what," "where," and "when" of episodes from a person's life [1]. The clinical 10 relevance of this episodic memory has only grown since then. For example, one of the current key 11 criteria in making a diagnosis of typical Alzheimer's disease (AD) is "an initial progressive and 12 predominant episodic memory deficit," [2]. In neuropsychological practice, episodic memory is 13 commonly assessed through tasks that sample different retrieval formats, including free recall, cued 14 recall, and recognition memory. Recognition memory testing is therefore one clinically important 15 way of probing episodic memory (although it does not capture it exhaustively and may reflect 16 different underlying processes). In routine assessment, clinical recognition memory testing is still 17 typically interpreted at the level of overall accuracy: whether the patient correctly endorses old items 18 and rejects new ones. That approach is useful, but it is often too blunt for early neurodegenerative 19 assessment. 20In this opinion piece, I argue that clinical recognition memory testing in older adults should move 21 beyond overall accuracy alone. Specifically, standard accuracy-based recognition measures should be 22 supplemented, where feasible, by brief subjective, process-sensitive methods that better distinguish 23 recollection from familiarity, both explained in greater detail later on. This is not simply a theoretical 24 preference but a clinically motivated proposal: standard delayed recognition memory can appear 25 relatively preserved in some older adults with amnestic mild cognitive impairment (aMCI), even 26 when recollection-based performance is already declining, and this may obscure diagnostically 27 meaningful change in prodromal AD [3][4][5][6][7][8][9][10]. 28Although the use of subjective judgments to investigate recognition memory processes is not new 29 and has long been central to research on episodic memory [11][12][13][14] term goal is not to introduce complex experimental paradigms into every assessment battery, but to 201 make modest changes to existing recognition phases (see Figure 1). 202 A feasible addition for routine practice is a brief remember/know judgment attached to recognised 203 items [41,[24][25][26][27][28]. After endorsing an item as old, the patient could be asked whether they 204 "remember/are sure of it" because specific details come back to mind, or whether they simply 205 "know/have a sense that" it was presented. This will not yield a process-pure estimate, but it can 206 provide clinically useful qualitative and semi-quantitative information if instructions are brief and 207 standardized. A second pragmatic option is to add confidence ratings to yes/no recognition decisions, 208 which can later be interpreted descriptively or, in research-oriented settings, incorporated into 209Receiver Operating Characteristic (ROC)-type analyses [20-21]. For example, patients are asked to 210 make yes/no judgements followed by confidence ratings for each response, even ones they said no to 211 (e.g., "on a scale of 1 (definitely old) to 6 (definitely new) do you recognise 'word x' being presented 212 before?"). 213Task formats can also be used strategically. The manner in which tasks are administered such as 214 forced-choice formats, picture-based recognition, time-limited decisions, frequency judgements, or 215Deleted: 40 The author declares that the research was conducted in the absence of any commercial or financial 303 relationships that could be construed as a potential conflict of interest. 304 The author declares that no funding was received to support the preparation of this opinion piece. 449