Semantic Priming Affects Immediate Serial Recall and Influences Order Errors

 SOUTHERN CROSS UNIVERSITY 

I have read and understand the Rules Relating to Awards (Rule 3 Section 18 – Academic Misconduct Including Plagiarism) as contained in the SCU Policy Library. I understand the penalties that apply for plagiarism and agree to be bound by these rules. The work I am submitting electronically is entirely my own work.

Learning and Memory (Session 1, 2020)

Abstract

Associative networks within long-term memory influence the working memory during tasks of immediate serial item recall. Semantically related word lists were remembered better than unrelated lists. These results were interpreted through semantic similarity effects, the reconstruction hypothesis, and the spreading activation model of semantic processing. This study further examined semantic priming effects in differently framed lists and found that related and unrelated cue words before encoding affected memory performance as well as item order. Thematically related lists that were cued with a semantically related word accounted for better recall than related lists primed with an unrelated cue. These priming effects were interpreted in terms of the embedded process framework, suppression and attention. Unrelated lists cued with an unrelated word performed worst; however, these lists created the least order errors. Order errors were highest in related lists with an unrelated primer. Order errors were discussed in terms of interference and suppression. Serial curves displayed a V-shape and were construed by an attentional gradient, subvocal rehearsal and semantic network interference with order errors shaping the curves.

Semantic Priming Affects Immediate Serial Recall and Influences Order Errors

From early childhood through to adulthood, we accumulate increasing amounts of knowledge, some of which we seem to accrue effortlessly, yet other learning involves long hours of study and rehearsing. We depend significantly on our memory capacities to acquire and remember movements, language, facts, and events. However crucial it is for many accomplishments in life, we still have gaps in our understanding of how long- term memory (LTM) and working memory (WM) engage synchronously.

LTM can hold large amounts of information, whereas WM is limited to 7 plus/minus two items or chunks of information in its capacity (Miller, 1956). The multi- component model of WM (Baddeley, 2012; Repovš & Baddeley, 2006) features the central executive that rapidly and fluidly manipulates information from the phonological loop and the visuospatial sketchpad. Encoding of perceptions relies on swift collaboration between auditory input and visual input, and cohesion relies on these representations held in the LTM. Alan Baddeley modified his theory by adding an episodic buffer to elucidate this intricate interaction between LTM and WM. WM performance on verbal memory tasks differs depending on the meaningfulness of the content already in LTM. The phonological loop, in conjunction with the visuospatial sketchpad, rehearsal, or chunking alone, cannot explain why semantically related items are remembered better in an immediate serial word recall task than unrelated words. A sense of coherence must involve LTM in conjunction with WM, just how it does this is not entirely understood.

This study aims to gain a deeper comprehension of how semantic memory stored in LTM influences the WM. Specifically, we are looking at priming effects that might help or hinder a participant from preparing for correct encoding and recall of a list of words within a semantically related or unrelated condition (RR-, UR- and UU- condition). Additionally, we are interested in serial recall and item order errors. Five hypotheses will shed further light on how semantic networks operate and how associations from LTM interfere with the reconstruction of a set of words depending on their categorical associations.

Provided that the brain is encoding information in the LTM within a semantic network (McCelland & Rumelhart, 1986), and according to the ‘spreading of activation model’ (Collins & Loftus, 1975), associable words would be memorized easier and recalled better than words that are not semantically related. This ‘semantic similarity effect’ (Monnier & Bonthoux, 2011; Saint-Aubin & Poirier, 1999; Tse, 2009: Tse et al., 2011) has been demonstrated in previous studies. Our study aims to replicate this effect with word lists containing thematically related and unrelated items. According to the ‘reconstruction hypothesis’ (Saint-Aubin et al., 2005), thematically encoded associated words are reconstructed from within a constrained frame and do not need to be drawn from distant, separate themes. Consequently, recall of items will be enhanced in word lists from the RR- and UR-condition, and inferior in lists from the UU-condition (H1). Semantic relatedness additionally acts as a primer for items on the lists.

How accessible a word is during recall depends on the priming that occurred before the encoding process. Tse et al. (2011) have found that there are differences in recall between categorically related lists and associatively related lists. The formation of semantic links within the network is formed over time. It does not necessarily evolve in categories but collaterally in combined experiences within a thematic world that is continuously updated. The ‘embedded processing framework’ (Cowan, 1988; Miyake & Shah, 1999) elucidates that a primer works like a stimulus that engages features from the LTM and additionally turns awareness and attention towards previous representations of an association. Accordingly, the RR-condition that is preceded by an associable primer will perform better than the UR- and the RR-condition that is preceded by an unrelated cue word (H2). Testing the recall of non-words compared to real words confirms, according to the ‘retrieval-based hypothesis’, that phonological representations help to reconstruct degrading memory traces (Saint-Aubin & Poirier, 2000).

When Ebbinghaus (1885/1964) experimented with non-words, he found that they were difficult to recall, he further noticed a U-shaped recall curve. Murdock (1962) studied different lengths of lists to examine the recall progression. He also noted a steep decline in retention and a slight recovery for the last words. Oberauer (2003) attributes this primacy and recency effect to a combination of interference and attentional gradient. Attention is allocated to the first items then gradually diminishes; at the same time, interference by following words overwrites the encoding. Consequently, our study results will see a U-shaped serial recall curve with a primacy and a recency effect (H3).

Semantically activated networks assist better recall of related word lists, yet according to Poirier et al. (2015), they are also responsible for a higher proportion or order errors in related word lists compared to unrelated lists. As the reconstruction of words within thematic frames is more accessible, the order of those items can get confused. Shared semantic features from the LTM affect ordering features of WM at recall. Hence, we expect a higher proportion of order errors in the RR- and UR- condition than in the UU-condition (H4). Alternatively, input and output interference (Oberauer, 2003) could account for order errors, as temporal coherence is overridden during either encoding or output.

The most intriguing aspect of our study is to examine the difference in order errors between the RR- and the UR-condition. If LTM interference affects order recall, then the need for suppression of the wrong category primer will increase encoding interference, and therefore order errors. Oberauer (2003) tried response suppression as an explanation for the recency effect. However, it is more likely that response suppression will affect order errors. Consequently, the UR-condition will accumulate more order errors than the RR-condition (H5).

Methods

Participants

The participants were 43 students at Southern Cross University enrolled in a unit concerned with learning and memory. There were 30 female, 10 male and 3 no sex declared participants and the mean age was 35.61 (SD = 12.13). English was the first language of 39 of the participants.

Materials and Apparatus

The task information and stimuli to be memorized were embedded in a PowerPoint presentation. The word stimuli appeared in the centre of a white screen in black, 96-point, Arial Black script. The stimuli consisted of 45 lists of 6 words with each list preceded by a category label (category prime) that was either related or unrelated to the words in the list: 15 lists of weakly semantically related words with a related category label (RR); 15 lists of weakly semantically related words with an unrelated category label (UR); 15 lists of semantically unrelated words with an unrelated category label (UU). The lists in the unrelated condition were made up from words contained in the other two lists in order to equate lexical factors (e.g.

Concreteness and Familiarity) amongst conditions. Finally, demographic information such as age, sex and whether English was their first language was recorded.  

Procedure

The experiment took place in a location of the participants choosing and they initially recorded their responses using pen and paper. Participants began by completing distractor tasks which were included to divert attention and minimise the participants guessing the main purpose of the study. Participants were then given a set of instructions indicating that they would see a series of word lists, with six words in each list. Each word was presented one at a time for 2 seconds. When the participants saw “*****”, this would indicate they should write down on a response sheet the words they recalled in the order the words were presented. If they could not remember a word, they were to make a mark in its place and go on to the next word. They could not backtrack or make corrections. When they had finished writing each list, they were to hit the space bar to begin the next list. After an initial practice trial, the 15 lists were presented in a mixed order. Finally, the participants were asked to score and transcribe their data onto an excel data collation sheet, including demographic questions, and return these to the experimenter.

Results

Our analysis focussed on the number of items recalled in their correct serial position, overall free recall, recall gradients from the first to the last item, and order errors within the three different conditions.

Recall performance was overall best in the RR-condition (total serial recall M = 4.36, SD = 0.80; total free recall M = 4.91, SD = 0.36), second in the UR-condition (total serial recall M = 4.10, SD = 0.82; total free recall M = 4.69, SD = 0.66) and worst in the UU-condition (total serial recall M = 3.84, SD = 0.75; total free recall M = 4.33, SD = 0.67), as visualized in Figure 1.

Figure 1. Performance over the different conditions overall with words recalled in total serial order and free recall words.

Individual item recall is illustrated in Figure 2.: The retention curves are indicating strong memory retention for the first word (RR = 93%, UR = 90%, UU = 87%), then gradually falling over the next four words, the fifth word performing poorest (RR = 57%, UR = 50%, UU = 46%) and a slight improvement on the last word (RR = 64%, UR = 56%, UU = 54%). The serial curves additionally recapture the overall performance of the three conditions with the RR-condition performing best and the UU- condition performing worst.

Figure 2. Serial recall of individual words (from word 1 to 6 in their corresponding lists) across the three different conditions.

The UR-condition was found to produce the highest proportion of order errors (M = 0.59, SD = 0.36), closely followed by the RR-condition (M = 0.55, SD = 0.36). The smallest proportion of order errors was recorded for the UU-condition (M = 0.49, SD = 0.33), as shown in Figure 3.

Figure 3. The proportion of order errors in the three different conditions.

Discussion

Our study addressed different aspects and theories of WM by manipulating semantic similarities to bring about a further understanding of how these theories interact and to generate a holistic view of the influence of LTM on WM. Our sample was very homogenous, mostly female university students with English as their first language. The experiment was not monitored in a lab, potentially participants could have made errors filing their data, or manipulated their test results. We have not instructed or asked about strategies used. Nevertheless, in line with previous immediate serial recall tasks (Monnier & Bonthoux, 2011; Saint-Aubin & Poirier, 2000; Saint- Aubin et al., 2005; Tse, 2009; Tse et al., 2011) semantically related word lists were recalled better than unrelated lists. This result confirms our first hypothesis, memory recall in the RR- and UR-condition outperformed the UU-condition. Furthermore, a notable difference was apparent between the RR- and the UR-condition, confirming our second hypothesis. Lists with a matching associative word primer were remembered better than lists with an unrelated primer. Drawing from Hayes ‘relational frame theory’ (Hayes & Bisset, 1998), associations are formed spontaneously also between B and C, not only the learned A to B and A to C associations. A spreading network of semantic associations (Collins & Loftus, 1975), activated by a primer, that tunes selective attention to a lexical field (Cowan, 1988), thus limiting the search scope during retrieval-based output (Saint-Aubin & Poirier, 2000), could explain this result obtained in our study. Examination of the serial recall curve displayed a V-shape rather than a U- shape, a deviation from our third hypothesis. Subvocal rehearsal (Baddeley, 2012) holds an explanation for the primacy effect; the first words in a list are rehearsed more often. They endure longer before decaying over time. Decay might explain the recency effect, as the last word had no time to decay, only needed to withstand the interference from recalling the previous words. The V-shape could be due to interference and decreasing attention allocated during the encoding of words (Oberauer, 2003), in combination with rehearsal loops (Baddeley & Hitch, 2019). However, previous order error studies (Poirier et al., 2015) have noted that in semantically related lists, words often appear in the wrong order. Notably, these order error charts displayed an inverted V-shape; hence this could quite clearly influence the shapes of our curves. Unfortunately, we were not able to distinguish in our analysis where exactly order errors appeared on an individual level. However, we have noticed more order errors in the related conditions than in the unrelated condition, confirming our fourth hypothesis and in line with previous studies (Saint-Aubin & Poirier, 1999; Poirier et al., 2015). While semantic networks support memory during immediate recall, their lexical activation in LTM allows for confusion in order information. However, higher recall in semantically related lists plausibly increases the probability of order errors (Saint-Aubin & Poirier, 1999; Saint-Aubin & Poirier, 2000), because more opportunity to confuse words exists rather than when fewer words were remembered overall. Notably, interference via an unrelated priming word caused additional confusion, as assumed in the fifth hypothesis. On the one hand, it impaired recall of words, and the semantical relatedness caused further order errors. As the primer cue word needed to be suppressed for successful encoding of the words from the list, this interference increased the chance for order errors (Oberauer, 2003).

Admittedly, the difference between order errors in the RR- and UR-condition was very small yet posed an exciting future avenue of exploration. If priming has an effect on order errors within semantically related conditions, is this due to additional suppression or interference? How do order errors influence the overall serial recall curve? It would further be interesting to examine how different cultural backgrounds and bilingualism affect semantic networks, associations and memory performance.

Consistent performance of verbal working memory relies on swift interactions between phonological and semantic encoding within pre-existing and coherent LTM networks. These associations further empower performance at recall. Semantic priming assists both encoding and recall. However, semantic priming by an unrelated word disrupts these processes and leads to errors in recall as well as increased order errors during recall.

References

Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual

review of psychology, 63, 1-29. https://doi.org/10.1146/annurev-psych-

120710-100422
Baddeley, A. D., & Hitch, G. J. (2019). The phonological loop as a buffer store: An

update. Cortex, 112, 91-106. https://doi.org/10.1016/j.cortex.2018.05.015 Collins, A. M., & Loftus, E. F. (1975). A spreading-activation theory of semantic

processing. Psychological review, 82(6), 407. https://doi.org/10.1037/0033-

295x.82.6.407
Cowan, N. (1988). Evolving conceptions of memory storage, selective attention, and

their mutual constraints within the human information-processing system. Psychological bulletin, 104(2), 163. https://doi.org/10.1037/0033- 2909.104.2.163

Hayes, S. C., & Bissett, R. T. (1998). Derived stimulus relations produce mediated and episodic priming. The Psychological Record, 48(4), 617-630. https://doi.org/10.1007/bf03395293

McClelland, J. L., Rumelhart, D. E., & PDP Research Group. (1986). Parallel distributed processing. Explorations in the Microstructure of Cognition, 2, 216-271.

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological review, 63(2), 81. https://doi.org/10.1037/0033-295x.101.2.343

Miyake, A., & Shah, P. (Eds.). (1999). Models of working memory: Mechanisms of active maintenance and executive control. Cambridge University Press.

Monnier, C., & Bonthoux, F. (2011). The semantic-similarity effect in children: Influence of long-term knowledge on verbal short-term memory. British Journal Of Developmental Psychology, 29(4), 929-941. https://doi.org/10.1111/j.2044-835x.2010.02024.x

Murdock Jr, B. B. (1962). The serial position effect of free recall. Journal of experimental psychology, 64(5), 482. https://doi.org/10.1037/h0045106

Oberauer, K. (2003). Understanding serial position curves in short-term recognition and recall. Journal of Memory and Language, 49(4), 469-483. https://doi.org/10.1016/s0749-596x(03)00080-9

Poirier, M., Saint-Aubin, J., Mair, A., Tehan, G., & Tolan, A. (2015). Order recall in verbal short-term memory: The role of semantic networks. Memory & Cognition, 43(3), 489-499. https://doi.org/10.3758/s13421-014-0470-6

Repovš, G., & Baddeley, A. (2006). The multi-component model of working memory: Explorations in experimental cognitive psychology. Neuroscience, 139(1), 5-21. https://doi.org/10.1016/j.neuroscience.2005.12.061

Saint-Aubin, J., & Poirier, M. (1999). The influence of long-term memory factors on immediate serial recall: An item and order analysis. International Journal of Psychology, 34(5-6), 347-352. https://doi.org/10.1080/002075999399675

Saint-Aubin, J., & Poirier, M. (2000). Immediate serial recall of words and nonwords: Tests of the retrieval-based hypothesis. Psychonomic Bulletin & Review, 7(2), 332-340. https://doi.org/10.3758/bf03212990

Saint-Aubin, J., Ouellette, D., & Poirier, M. (2005). Semantic similarity and immediate serial recall: Is there an effect on all trials. Psychonomic Bulletin & Review, 12(1), 171-177. https://doi.org/10.3758/bf03196364

Tse, C. (2009). The role of associative strength in the semantic relatedness effect on immediate serial recall. Memory, 17(8), 874-891. https://doi.org/10.1080/09658210903376250

Tse, C. S., Li, Y., & Altarriba, J. (2011). The effect of semantic relatedness on immediate serial recall and serial recognition. The Quarterly Journal of Experimental Psychology, 64(12), 2425-2437. https://doi.org/10.1080/17470218.2011.604787