Adam F. Osth

Sources of interference in item and associative recognition memory.

A powerful theoretical framework for exploring recognition memory is the global matching framework, in which a cues memory strength reflects the similarity of the retrieval cues being matched against the contents of memory simultaneously. Contributions at retrieval can be categorized as matches and mismatches to the item and context cues, including the self match (match on item and context), item noise (match on context, mismatch on item), context noise (match on item, mismatch on context), and background noise (mismatch on item and context). We present a model that directly parameterizes the matches and mismatches to the item and context cues, which enables estimation of the magnitude of each interference contribution (item noise, context noise, and background noise). The model was fit within a hierarchical Bayesian framework to 10 recognition memory datasets that use manipulations of strength, list length, list strength, word frequency, study-test delay, and stimulus class in item and associative recognition. Estimates of the model parameters revealed at most a small contribution of item noise that varies by stimulus class, with virtually no item noise for single words and scenes. Despite the unpopularity of background noise in recognition memory models, background noise estimates dominated at retrieval across nearly all stimulus classes with the exception of high frequency words, which exhibited equivalent levels of context noise and background noise. These parameter estimates suggest that the majority of interference in recognition memory stems from experiences acquired before the learning episode.

Stimulus type and the list strength paradigm

In recognition memory, increasing the strength of studied items does not reduce performance on other items, an effect dubbed the null list strength effect (LSE). While this finding has been replicated many times, it has rarely been tested using stimuli other than single words. Kinnell and Dennis (2012) recently tested for the presence of list length effects using non-word stimulus classes while controlling for the confounds that are present in list length designs. Small list length effects were found for fractal and face images. We adopted the same paradigm and stimuli used by Kinnell and Dennis to test whether these stimuli would be susceptible to list strength effects as well. We found significant LSEs for fractal images, but null LSEs for face images and natural scene photographs. Stimuli other than words do appear to be susceptible to list strength effects, but these effects are small and restricted to particular stimulus classes, as is the case in list length designs. Models of memory may be able to address differences between these stimulus classes by attributing differences in representational overlap between the stimulus classes.

Associative recognition and the list strength paradigm

When a subset of list items is strengthened, the discriminability of the nonstrengthened items is unaffected. This regularity has been dubbed the null list strength effect (LSE), and despite its many replications in item recognition, little research has investigated whether an LSE occurs in associative recognition. We conducted two experiments in which a set of pairs were studied once and a set of interference pairs were studied either once (pure-weak-list condition) or four times (mixed-list condition). Equivalent levels of performance for the nonstrengthened pairs were observed in both the pure-weak and mixed conditions using both yes–no and two-alternative forced choice testing. Additionally, equivalent false alarm rates were observed between rearranged pairs composed of weak and strong items. Both sets of results were found to be consistent with a matrix model that has no overlap among its item representations.

Modeling the dynamics of recognition memory testing with an integrated model of retrieval and decision making

A robust finding in recognition memory is that performance declines monotonically across test trials. Despite the prevalence of this decline, there is a lack of consensus on the mechanism responsible. Three hypotheses have been put forward: (1) interference is caused by learning of test items (2) the test items cause a shift in the context representation used to cue memory and (3) participants change their speed-accuracy thresholds through the course of testing. We implemented all three possibilities in a combined model of recognition memory and decision making, which inherits the memory retrieval elements of the Osth and Dennis (2015) model and uses the diffusion decision model (DDM: Ratcliff, 1978) to generate choice and response times. We applied the model to four datasets that represent three challenges, the findings that: (1) the number of test items plays a larger role in determining performance than the number of studied items, (2) performance decreases less for strong items than weak items in pure lists but not in mixed lists, and (3) lexical decision trials interspersed between recognition test trials do not increase the rate at which performance declines. Analysis of the models parameter estimates suggests that item interference plays a weak role in explaining the effects of recognition testing, while context drift plays a very large role. These results are consistent with prior work showing a weak role for item noise in recognition memory and that retrieval is a strong cause of context change in episodic memory.

A diffusion decision model analysis of evidence variability in the lexical decision task

The lexical-decision task is among the most commonly used paradigms in psycholinguistics. In both the signal-detection theory and Diffusion Decision Model (DDM; Ratcliff, Gomez, & McKoon, Psychological Review, 111, 159–182, 2004) frameworks, lexical-decisions are based on a continuous source of word-likeness evidence for both words and non-words. The Retrieving Effectively from Memory model of Lexical-Decision (REM–LD; Wagenmakers et al., Cognitive Psychology, 48(3), 332–367, 2004) provides a comprehensive explanation of lexical-decision data and makes the prediction that word-likeness evidence is more variable for words than non-words and that higher frequency words are more variable than lower frequency words. To test these predictions, we analyzed five lexical-decision data sets with the DDM. For all data sets, drift-rate variability changed across word frequency and non-word conditions. For the most part, REM–LD’s predictions about the ordering of evidence variability across stimuli in the lexical-decision task were confirmed.

Prior-list intrusions in serial recall are positional.

Henson (1996) provided a number of demonstrations of error patterns in serial recall that contradict chaining models. One such error pattern concerned when participants make intrusions from prior lists: Rather than originating from random positions in the prior list, intrusions tend to be recalled in the same position as their position in the prior list, a finding which led to the endorsement of positional models of serial recall. However, all of the demonstrations of positional intrusions occurred in designs in which relatively small sets of items were repeatedly employed as stimuli. In recent years, a number of investigations have found evidence for chaining in designs in which large sets of items are employed and items are never reused across trials (open sets). We conducted 2 experiments using open sets of items to test whether a pure chaining model is a viable model for open-set conditions. Both experiments revealed that intrusions from the immediately preceding list exhibited a strong tendency to be output in the same position as their position in the prior list, suggesting the usage of positional representations in open-set designs. A chaining model that lacks positional representations provides an inadequate account of serial recall in open-set conditions.

Likelihood-Free Algorithms

In this chapter, we will present several algorithms, which differ in how they approximate the likelihood function and generate proposals for the posterior distribution, for performing likelihood-free inference. Four classes of algorithms—rejection-based, kernel-based, general methods, and hierarchical—will be discussed in great detail. We will provide a brief overview of the origins of each class as well as discussing the advantages and disadvantages of each. Finally, we will close the discussion by offering guidance on how to choose the appropriate class of algorithms for use in a given situation.

The fill-in effect in serial recall can be obscured by omission errors.

Henson (1996) provided a number of demonstrations of error patterns in serial recall that contradict chaining models. Chaining models predict that when participants erroneously recall an item too early, recall should proceed from the point of error. In contradiction to such a prediction, Henson found evidence for a fill-in effect: participants were much more likely to revisit an erroneously skipped item than to continue onward to later list items. However, recent reanalyses of serial recall data sets have found evidence for the opposite pattern in serial recall experiments that use open sets of items. We tested the hypothesis that open sets of items produce fill-in effects by comparing serial recall with an open set and a closed set, and when participants were allowed and prohibited from skipping over responses, and comparing serial recall with a reconstruction of order task. Fill-in effects were observed in all cases except when participants were not encouraged to skip over responses. Subsequent analyses indicated that when omission rates were equated, a fill-in effect was observed for all conditions when lists contained no omissions. These results suggest that high omission rates in open-set designs obscure a fill-in effect and further sound a cautionary note about interpreting cases in which recall continues in the forward direction after a skipped response.