Healthy aging is accompanied by cognitive alterations, in addition to physical and neurobiological changes . For instance, cognitive processes such as processing speed, working memory, episodic memory and executive functions, are susceptible to aging-related decline, while other cognitive abilities like semantic memory and problem-solving are preserved or even improve over the life span [1,2,3,4,5]. Moreover, following high-level education early in life, midlife work complexity, and late-life active engagement in social, physical and mental activities may all have a protective effect on aging-related cognitive disorders . Successful participation in later-life activities and optimal functioning in a rapidly changing daily life thus requires optimal and life-long learning.
One approach that may improve learning outcome in healthy older adults (OA) is errorless learning (EL). Core element of the EL approach is that the occurrence of errors is prevented as much as possible (or even fully eliminated), typically resulting in superior learning compared to trial and error learning (TEL) in cognitively unimpaired older adults [7, 8]. However, beneficial effects of TEL, in which errors during learning are not eliminated, have also been reported in OA . In contrast, others did not find any differences in performance in OA between EL and TEL [but showed a superior effect after EL in healthy young adults (YA)] . Based on these mixed findings in healthy YA and OA, it has been argued that age may not be the defining factor, but that the adverse or beneficial effect of errors depends on the type of information to be encoded and retrieved. Cyr and Anderson, for instance, suggested that both YA and OA would benefit from errors in conceptual learning, whereas EL is preferable in nonconceptual learning [11, 12]. To either benefit from or being hampered by errors made during learning may also rely on the ability to successfully distinguish between correct and erroneous responses (that is, error monitoring, part of cognitive control). Several studies found that error monitoring declines with increasing age, resulting in slowing responses and greater proportion of undetected errors compared to YA [13,14,15].
To date, very little is known about the impact that the number of errors committed (i.e., error frequency) during acquisition has in healthy YA and OA. The role of error frequency has often been overlooked in previous studies, as paradigms neither employed systematical manipulation nor analyzed error rates. Note that in most commonly used EL paradigms participants had to complete word stems or word pairs for which, in the EL condition, correct answers were immediately presented by the experimenter or, in the TEL condition, correct answers had to be guessed by the participants in two to four attempts, but that were eventually also provided by the experimenter. This lack of systematic manipulation or control complicates the interpretation of previous studies may be a driving force behind inconsistent results .
In the present study, we compared memory recall performance in healthy YA and OA after EL and TEL using a spatial learning task, in which the amount of errors (defined as incorrect responses) committed during the acquisition phase of TEL was carefully manipulated . That is, participants made 0, 2, 3, 4, or 5 incorrect responses (i.e., errors) before their response was considered to be correct. Moreover, as remembering word pairs or completing word stems bears little resemblance to everyday life demands, we adopted a more ecologically valid approach, that is, by employing a visuospatial learning task in which participants had to search for objects at different locations and remember these for later use, processes that are also highly relevant for everyday functioning (e.g. enabling us to recall where we have stored our wallet, keys, or glasses) . We hypothesized that YA would perform better after EL compared to TEL, but that error frequency would not influence recall performance in this group, thus replicating previous findings in YA . In addition, Kessels et al.  and Ariel and Moffat  demonstrated that OA perform worse than YA on explicit spatial learning and memory tasks, for instance tasks in which participants had to acquire, store and retrieve the locations of everyday objects in one of five virtual rooms (living room, bedroom, study room, bathroom and kitchen), but that implicit spatial learning, metacognition and navigation were largely preserved at older age. Based on these findings and the finding that error monitoring is susceptible to aging-related decline [14, 15], thus adding to the effects of ageing-related episodic memory decline [1,2,3,4], we hypothesized that a higher amount of errors would interfere more with learning in OA than in YA, which should be reflected by a reduced recall accuracy in the OA specifically.