Memory for serial order is essential for the management of high-level cognitive activities [1], influencing the ability to recall information independently of the position of the items. Cognitive studies have extensively explored the problem of serial order in different domains, employing different methodological procedures, and they repeatedly reported evidence consistent with the serial position effect. The serial position effect reflects the systematic changes in accuracy across an item’s position, showing a significantly higher performance when responding to the first – primacy effect – and to the last items – recency effect – of a sequence [2], whereas the middle items tend to be forgotten. The presence of the serial position effect is typically displayed by a U–shape curve when plotting the recall accuracy as a function of the serial order of the items [1].

An extensive amount of evidence has demonstrated the occurrence of the serial position effect in different situations involving the employment of verbal materials, such as letters and words [3-5]. According to Hurlstone et al. [1], the main use of verbal material in research dealing with the serial position effect may be due to the easy wayof handling, that is, building, manipulating and testing verbal material compared to non–verbal material. On the other hand, it is possible that memory functions are sensitive to the stimulus domain [6], showing different results in the non–verbal domain. As for non–verbal material, there are controversial results in studies using visuospatial stimuli.

Comparing verbal and visuospatial material has not provided well-established and consistent results. Indeed, Cortis et al. [7] found the same effect of serial order for both verbal and visuospatial material, claiming that different results across modalities found in previous studies could depend on the not–uniform methods used [8-10]. Thus, their findings seem in line with the assumption of Guérard and Tremblay [11], who proposed a functional equivalence of serial recall, suggesting that the serial order may be elaborated in similar ways across the domains [1]. However, despite providing encouraging results, some critical points in previous studies should be considered. In a study by Cortis et al. [7], the overall accuracy differed between verbal and visuospatial stimuli, and primacy and recency effects were weaker for visuospatial than for verbal stimuli. Moreover, Hurlstone et al. [1] highlighted that the serial position effect is well–established in the verbal domain. At the same time, further investigation is necessary for the visuospatial domain due to not totally clear results. Thus, it seems that the serial position effect is more pronounced in the verbal domain compared to the visuospatial domain [12].

It is interesting to note that in the visuospatial domain, different types of material were used to test the serial order influence, such as sequences of visuospatial locations [9, 13, 14], visuospatial movements [15, 16], and routes navigation [17, 18]. A multitude of different stimuli employed, but to the best of our knowledge; no study has investigated whether a verbal–spatial type of stimuli would determine the same pattern of results. In particular, there is no evidence clarifying whether the memory recall of items in a spatial description (i.e., verbal description of an environment) is affected by the serial order differently from the recall of items in a classic word list.

Previous evidence in spatial literature suggests that an environment verbally described is spatially, and not textually, encoded and fosters the development of a mental representation with the spatial characteristics described [19-21]. Moreover, spatial descriptions seem to be functionally equivalent to directly perceived scenes since they preserve metric information and structural coherence [22]. Previous studies suggest that serial order is more bound to verbal than spatial information in working memory [12], and that spatial information is strategically chunked in local spatial configuration [23]. Consequently, when encoding a spatial description, we can expect two main scenarios. On the one hand, the spatial description is encoded as verbal information and will then rely on the serial position order; in this case, it would be exposed more easily to serial position effects. On the other hand, if the spatial description is encoded as spatial information, it will be unbound to serial order and more prone to other organization strategies. Thus, if the spatial description is encoded as spatial material, we should expect a response to serial order different from that typically shown by verbal stimuli.

Concreteness is an important factor that typically affects the recall of verbal items in a serial order. Empirical evidence has demonstrated that concreteness and other features of the language, such as word frequency and lexical status, affect the number of items successfully recalled [24]. According to the authors, concreteness seems to foster memory processes since representations of concrete items contain more information than those of abstract items. In particular, concreteness strengthens item memories and consequently leaves time and resources available to process serial orders. This effect can be interpreted according to Paivio’s dual coding theory [25], which assumes that concrete words maintain both a verbal and an imagistic code. Moreover, it is consistent with the idea that using different types of information sources leads to better performance [26]. The concreteness of the items might be enhanced by including them in a meaningful context; indeed, it has been demonstrated that items are better recalled in a meaningful context than in a list [27, 28]. According to Brodsky et al. [29], the logical structure which joins items seems to have a crucial role in organizing memory processes.

Based on our analysis, spatial coding and meaningful context are the two factors that might determine a different serial position effect for items encoded in a spatial description than in a word list. Indeed, the spatial description provides information relative to the spatial relations between the described objects. Moreover, the described objects are included in a meaningful context, enhancing the ability to visualize the described environment mentally. Therefore, both the spatial coding and meaningful context characterise the spatial description. In order to determine the “pure” influence of spatial coding on the serial position effect of items included in a spatial description, it is necessary to test the occurrence of the serial position effect on items described within a meaningful narrative, which provides no spatial information.

The present study aims to investigate the occurrence of the serial position effect in the recall of items verbally presented in three different contexts: a classic word list, a spatial description of a room and a narrative without spatial information. We expect different accuracy distributions across item positions for the three contexts. In particular, we hypothesise that the spatial description will be encoded as spatial material, reducing the influence of the serial order and consequently determining a flattened U–shaped curve. Moreover, we hypothesise that the accuracy distribution for the spatial description will be different from the accuracy distribution for the narrative without spatial information.

Seventy–five university students (M = 19; F = 56) participated in this experiment in exchange for academic credits. Their age varied from 19 to 51 (M = 22; SD = 4.9). All participants were native Italian speakers. All participants reported that their psychophysiological state was not affected by alcohol consumption or insufficient sleep in the last 24 hr [30]. The participants signed the informed consent before starting the experiment. Participants were naive as to the purpose of the experiment.

We employed an experimental design with two independent variables: Context (between subjects) and Position (within subjects). As regards Context, participants were randomly assigned to three conditions: List (L), Spatial (S), and Narrative (N). The Context variable refers to the context in which fifteen objects were verbally described in the learning phase. Indeed, in the List condition, participants listened to a list of fifteen objects; in the Spatial condition, participants listened to the description of a fictitious room containing the same fifteen objects described in the list; in the Narrative condition, participants listened to a meaningful narrative, in which the fifteen objects are described in a non–spatial context.

The Position variable refers to the position of the fifteen objects described in the three contexts. Indeed, the position of each object was kept unchanged across the three Context conditions. To effectively investigate the shape of the serial position effect, the variable position was rendered discrete with three levels. The fifteen objects were grouped into five clusters of three objects each. Then, the Initial condition refers to the first three objects (1–3), the Central condition refers to the three central objects (7–9), and the Final condition refers to the final three objects (13–15).

To provide participants with auditory information (narratives description and testing trials), we employed a notebook connected with Sennheiser HD515 headphones. The same notebook, running E–Prime 2 Software, was used to generate trials and perform the task.

We chose fifteen words that were comparable in terms of both frequency use in the Italian language and the number of letters. According to the experimental design, we manipulated the context in which the fifteen objects were presented. In the List condition, the fifteen words were presented in a sequence, as for example, “carpet, pillow, backpack, […]”. Differently, in the other two Context conditions, the same words were included in a verbal description. Specifically, in the Spatial condition, the words were included in the spatial description of a room. This kind of description included spatial coordinates for each object and the spatial relation between two subsequential objects, for example “at the right corner, above the carpet on the floor, there is a pillow and a backpack [...]”. Conversely, in the Narrative condition, the words were part of a story. This story described the objects in the room concerning their role in the narrative, rather than to spatial information, as for example, “I lie down on the carpet, leaning my head on the pillow and facing the backpack [...]”. The total number of words in the spatial description and narrative was analogous. In a pilot study, we tested the appropriateness of the descriptions provided. In particular, we tested the text comprehension difficulty in the spatial description and narrative; moreover, only for the spatial description, we asked participants to evaluate the ease in mentally representing the room described. The results confirmed the appropriateness of the description provided ¹.

Both the list of words and the verbal descriptions were read by an experimenter and recorded. The time between two subsequent words was comparable across the three Context conditions to avoid any possible confounding effect of time.

The experimental procedure consisted of learning, in which, depending on the condition, participants listened to the context (either in the list or in the descriptions), and a testing phase, in which participants performed an old/new recognition task.

Before beginning the learning phase, participants read and signed the informed consent. Then they were accompanied into a silent, dimly lit room, positioned comfortably on a chair in front of a computer, and asked to wear the headphones. The duration of the learning phase differed across the Context conditions, since participants were exposed to the fifteen objects described within three different contexts. In the List context, they listened to a serial list of the objects; in the Spatial context, they listened to a verbal description of objects with precise spatial coordinates; in the Narrative context, they listened to a story that featured each object in a narrative. Before listening to the contexts, in all the conditions, participants were asked to listen to the stimuli carefully and to memorize them. No information was provided to participants as regards the subsequent task.

The testing phase started immediately after the learning phase. As soon as the list or the descriptions ended, participants were asked to observe the monitor and read the instructions explaining the following old/new recognition task. The task required participants to decide whether the acoustically provided words were old or new by pressing two alternative keys on the keyboard. There were thirty words: fifteen of them were the names of new objects, and the other fifteen were the names of the objects previously heard in the learning phase. Participants were exposed to three repetitions of the words in random order. After each repetition, participants were allowed to take a little break. Both accuracy and response times were measured.

Fig. (1). Distribution of accuracy scores across item positions for the list (a), spatial (b) and narrative (c) conditions. Error bars indicate standard errors.

Since the experimental procedure combined different types of stimuli (i.e., word lists and verbal descriptions), we could not employ the test procedures commonly adopted with one of the two types of stimuli, such as free or serial recall [31, 32] and story retell procedure 29, respectively. Free or serial recall would require participants to autonomously report all the words they can recall in a list format. This procedure would be adequate to measure accuracy in the List condition but could promote memorization if the items are in a list form, potentially suppressing any effect determined by the Spatial and Narrative contexts. Similarly, a story retelling procedure would be suited in the case of the Narrative condition, but it could as well promote a style of memorization that could suppress the effect of the Spatial and List contexts. To sum up, none of these procedures could be adequately applied to all three contexts. Therefore, to expose participants to the same test procedure in all conditions, we decided to employ a two alternative forced choice task [33-36].

As regards accuracy, we performed a 3 x 3 (Context x Position) repeated measures ANOVA, revealing a significant main effect of Context [F(2, 72) = 11.912; p < .001; η2 = .249], a main effect of Position, [F(2, 144) = 3.505; p < .05; η2 = .046], and a significant interaction, [F(4,144) = 5.658; p < .001; η2 = .136]. Planned contrasts with Bonferroni correction showed that participants were more accurate in the List and Spatial conditions than in the Narrative condition (p < .001 and p < .005, respectively), while the comparison between the List and Spatial conditions was only marginally significant (p = .09). Since the interaction reached a significant value, we ran further statistical analyses to understand the direction of the effect better; thus, we performed separate analyses for each Context condition (Fig. 1).

As regards the List condition, a repeated measures ANOVA for Position showed a significant main effect, [F(2,48) = 10.012; p < .001; η2 = .294]. Planned contrasts with Bonferroni correction revealed that participants were more accurate in the recognition of words in both the Initial and Final conditions than in the Central condition (p < .001 and p < .05, respectively), while no difference emerged between the Initial and Final conditions. Moreover, the statistics showed a significant quadratic trend, [F(1,24) = 17.155; p < .001; η2 =.417].

As regards the Spatial condition, we performed repeated measures ANOVA for Position, which revealed neither a significant main effect nor a significant quadratic trend.

Finally, in the Narrative condition, we performed a repeated measures ANOVA for Position, finding a significant main effect, [F(2,48) = 4.103; p < .05; η2 = .146]. Planned contrasts with Bonferroni correction revealed no difference either between the Initial and Central conditions or between the Central and Final conditions, while accuracy in the Initial condition was statistically lower than in the Final condition (p < .01). Moreover, no significant value was found for the quadratic trend. Instead the data revealed a significant linear trend, [F(1,24) = 8.048; p < .01; η2 = .251].

As regards response times, we performed a 3 x 3 (Context x Position) repeated measures ANOVA, which revealed neither significant main effects nor significant interaction.