Language Processes, Meaning/Semantics, Working Memory

Never Eat Sour Watermelons: Remembering lists as sentences

Author ,
Other Posts By ,
No Comments

Everyone struggles to remember things. Whether it’s a home address, someone’s phone number, or even a new acquaintance’s name—it’s challenging to recall information when we need to. We have lost count of the number of times we’ve been introduced to someone only to immediately forget their name. However, some things might be easier to remember than others. This could be due to how our working memory interacts with our knowledge of language.

For example, take the common mnemonic to remember the cardinal direction (NESW): Never Eat Sour Watermelons. Or, perhaps, ‘Never Eat Shredded Wheat’ or even ‘Never Eat Soggy Waffles.’ Who doesn’t love a good food-themed mnemonic? Why might these silly phrases be easier to retain than if you mixed up the order of the words to, “Watermelons Never Eat Sour?” Technically both phrases are correct—the order of the directions is the same in both cases—but remembering the second phrase is much more difficult. This intuition encourages us, and the authors of a recent paper published in Memory & Cognition, to ask, “What makes some phrases so memorable?”

A black and white photo of a herd of sheep standing shoulder-to-shoulder. Recalling something from memory can be like trying to herd sheep
Verbal working memory is a complex cognitive phenomenon whose relationship with long-term memory is still debated. One of the big questions is whether verbal working memory is a separate storage, or ‘buffer,’ or if it merely highlights information in long-term memory. Photo from Pixabay.com

Verbal working memory is commonly conceptualized as a temporary store of language-based information (e.g., addresses, lists, names, etc.). What remains mysterious is the relationship between verbal working memory and long-term memory. Buffer theories of working memory posit that there is an abstracted, and separate, store of knowledge for what is currently in mind which has limited contact with long-term memory.

On the other hand, emergent theories suggest that verbal working memory is not distinct from long-term memory but is rather what is currently under the attentional spotlight. To help differentiate the two, imagine a field full of sheep. Buffer theories are like if a herding dog separated some of the sheep into a temporary pen—separate from the rest of the herd. Emergent theories are more like a helicopter pointing a floodlight down onto a small section of the herd from above. The herd is intact, but some are highlighted. The question remains, which theory best captures how verbal working memory works?

Pictures for three smiling faces of the four authors of the current paper.
Authors of the featured article “Lexico-syntactic constraints influence verbal working memory in sentence-like lists.” Steven Schwering (left), Cassandra Jacobs (middle), Janelle Montemayor (not shown), and Maryellen MacDonald (right).

Turns out, understanding why some phrases are more memorable than others and which verbal working memory theory aligns best with collected evidence can be done at the same time. Steven Schwering, Cassandra Jacobs, Janelle Montemayor, and Maryellen MacDonald (pictured above) set out to accomplish just that in their recent study.

The authors carefully created ‘sentence-like’ lists that mimic some of the structural regularities of sentences. For example, take the word list, ‘Large Waiter Helps Tall Soldier Not Old Bishop.’ Intriguingly, sentence-like lists have been shown to result in better accuracy during recall than typical lists. The question remains is why people are more likely to remember lists that are bound together by semantic and syntactic information. The key to adjudicating between the two dominant theories of verbal working memory is to see whether working memory receives a boost from general linguistic knowledge—a boost that strict buffer theories say shouldn’t exist.

In a series of three experiments, the authors manipulated the degree of ‘sentence-likeness’ lists of words by swapping out the specific words for different verbs or nouns. In this way, the six-word list might resemble a sentence with a subject, verb, direct object, and indirect object (see Figure 3 for how these different lists looked). In the first experiment, the third position was either a noun or a ditransitive verb (a verb like “give” that takes both a direct object and an indirect object).

A figure showing the processing of converting a sample dative sentence into a sample list. The example sentence is “The responsible chemist awarded the ridiculous director a chair.” Taking out articles produces the list for the verb condition in the experiment, which is “responsible chemist awarded ridiculous director chair.” In the noun condition, the authors swapped the verb of the list out for another noun, here producing “responsible chemist railing ridiculous director chair.” The diagram also shows that every other list followed this same pattern of adjective, animate noun, verb/noun, adjective, animate noun, and then inanimate noun.
Figure 2 from the featured article. Examples of the lists used in Experiment 1 of the study. In the verb condition, the third word in the list is a ditransitive verb (e.g., awarded), and in the noun condition, that position in the same list is another noun (e.g., railing).

In the first experiment, participants had a much easier time recalling the words in the exact order from before, indicating that participants might be using their knowledge of language to remember the lists as sentences, but some modified buffer accounts actually predict this too. These buffer accounts say that part-of-speech sequence information like subject-verb-object word order can be indexed in verbal working memory, so the authors continued to manipulate the sentence-like lists.

In the second experiment, the third position in the list was either a ditransitive verb or an intransitive verb (a verb that takes no objects like “sleep”). In the third experiment, the authors held the third position constant as a ditransitive verb, and instead manipulated the second position to be either an animate noun or an inanimate noun. Since inanimate nouns wouldn’t fit into a sentence with a ditransitive verb, it should lower how ‘sentence-like’ the list is.

Line graph of proportion correct versus position in the list showing the difference in memory performance for whether the third position in the list was a noun or a verb. The line representing the lists with verbs (i.e. sentence-like lists) have a higher proportion correct at each word position.
Figure 3 from the featured article, showing lines for performance in each condition. Participants were much more likely to remember any of the words from sentence-like lists than from random word lists.

In these next two experiments, the authors found small effects for recalling the words in the sentence-like lists with either a ditransitive verb or an animate noun. However, the differences between the two conditions are relatively subtle, and the authors interpreted their results as tentative support for emergent theories of long-term memory and verbal working memory. Some buffer theories do account for some effect of general linguistic knowledge, but the authors call for more research to fully differentiate between the two theories.

Two more line plots from the featured article showing the results of Experiment 2A and Experiment 2B. Just like the previous figure, the position of the word in the list is on the x-axis and the proportion of words recalled correctly is on the y-axis. The figure on the left has conditions for ditransitive and intransitive verbs. The figure on the right has conditions for inanimate and animate nouns. The lines in both figures follow each other closely, then diverge slightly after position three or four.
Strict serial scoring results from Experiments 2A and 2B, or Figures 6 (left) and 7 (right). In both experiments, there is a much smaller effect of sentence-likeness on word memory.

A striking part of the featured paper is just how subtle these experimental manipulations need to be in order to answer these important questions about fundamental cognitive processes. Linguistic knowledge and memory interact in important ways, and it’s critical that our theories can explain these interactions.

The authors remarked on the value of this study beyond the data, saying that looking

“outside the standard toolkit of working memory methods to adopt methods from language comprehension can reveal effects that wouldn’t be identified in more typical working memory studies.”

So, next time you need to remember your grocery list or some folks you met at a conference, try to leverage your knowledge of language to make your memory a little better.

Featured Psychonomic Society article

Schwering, S. C., Jacobs, C. L., Montemayor, J., & MacDonald, M. C. (2023). Lexico-syntactic constraints influence verbal working memory in sentence-like lists. Memory & Cognition, 1-19. https://doi.org/10.3758/s13421-023-01496-2

The Psychonomic Society (Society) is providing information in the Featured Content section of its website as a benefit and service in furtherance of the Society’s nonprofit and tax-exempt status. The Society does not exert editorial control over such materials, and any opinions expressed in the Featured Content articles are solely those of the individual authors and do not necessarily reflect the opinions or policies of the Society. The Society does not guarantee the accuracy of the content contained in the Featured Content portion of the website and specifically disclaims any and all liability for any claims or damages that result from reliance on such content by third parties.

You may also like