Physical and semantic space at encoding

Encoding in verbal memory involves the perception of words and generation of mental representations (Paivio, 1990). There is evidence that these representations do not only involve semantic information, but also information with regard to the location of words (see Spivey, Richardson, & Fitneva, 2004 for a review). Pylyshyn (1989) was among the first to propose a model (i.e., spatial indexing) in which locations are encoded along with the words themselves.

In this model, a pre-attentive or a pre-cognitive mechanism in the visual system indexes the location of a visual stimulus by separating locations from other visual features (e.g., colour, shape etc.) even before recognising visual patterns.

Abundant evidence indicates that indexed locations interact with locations denoted by words upon presentation. For example, people react faster and more accurately in naming tasks if a word that conveys specific and explicit spatial information such as “left” appears in a congruent location on the display (i.e., left side of the screen) (White, 1969). On the other hand, inconsistency between spatial representations results in slower and less accurate responses (see Lu & Proctor, 1995 for a review). Known as spatial Stroop, this effect occurs even though the information about the word location is not necessary to complete the task successfully.

Reading words, which do not denote a location explicitly such as “left” but imply implicit locations such as “bird” at certain locations, has also consequences on how well these words are processed. Zwaan and Yaxley (2003) evidenced facilitation with spatial congruency in the case of words with imagined locations such as “basement” rather than the locations in the subject’s immediate environment. Participants were tasked to identify whether “root” and

“branch” are semantically related. Reactions were faster and more accurate if “root” was located under “branch” compared to the incongruent arrangement (i.e., “branch” under “root”).

In a similar paradigm, participants in Šetić and Domijan (2007) were tasked to verify whether a given word refers to a flying or a non-flying animal. Words for flying animals were verified

faster when they were presented at the top of the screen and words for non-flying animals were verified faster when they were presented at the bottom (Experiment 1). The same effect was observed when participants verified living or non-living objects that are associated with upward or downward positions, with a categorisation task that was not spatial (Experiment 2).

Words that make direct reference to spatial locations (e.g., “above”) and words that occur in imagined but typical locations (e.g., “bird”) are not the only words associated with locations in space. Emotionally charged words with positive or negative valence such as “hero” or “liar”

activate spatial simulations as well based on metaphorical relationship between valence and vertical space (i.e., “good is up”, “bad is down”) (Lakoff & Johnson, 1980b). Drawing on the associations between emotional words and space, Meyer and Robinson (2004) demonstrated that positive words are evaluated faster (i.e., “Rate ‘hero’ on a scale from 1 (negative) to 5 (positive)) if they are presented at the top of the screen and negative words (e.g., “liar”) are evaluated faster if they are presented at the bottom. In a similar fashion, Giessner and Schubert (2007) showed that people evaluate leaders as more powerful if their picture are positioned higher in a vertical arrangement (see also Schubert, 2005). In the domain of time - space mapping, it was demonstrated that linguistic units that imply time (e.g., “yesterday”, “we will drive”, “next” etc.) were judged faster when their position (i.e., left or right of the screen) matched with the location of response button (i.e., left or right hand-side button on the keyboard). For example, participants were faster to evaluate whether the sentence “we will drive” refers to past or future if the sentence was presented on the right side of the screen and response button was on the right hand-side (Santiago et al., 2007).

These studies reveal two mechanisms regarding space and language: First, word locations cannot be ignored even when they are irrelevant to the task (see Hock & Egeth, 1970; Seymour, 1977). In other words, individuals encode word locations automatically along with the words (see Chapter 3). Second, reading words with implicit spatial meaning simulates these locations automatically. When these two automatic mechanisms take place at the same time, a

compatibility between the physical and sematic location of a word facilitates the response in tasks that require conceptual processing; whereas a conflict between these locations delays the processing.

Along with that, how simultaneous activation of physical and semantic representations of space affects memory performance is not clear. As a first question, we tested whether spatially congruent encoding (e.g., encoding “bird” at the top of the screen) results in faster retrieval in comparison to spatially incongruent encoding (e.g., encoding “bird” at the bottom of the screen).

The intuitive prediction in line with the studies discussed above is that spatial congruency at encoding leads to better memory. Under this prediction, participants are expected to process spatially congruent words faster and in contrast, spatially incongruent words slower. There is evidence to assume that encoding difficulty results in impaired memory as frequently seen in schizophrenia (M. J. Smith, Gorman, Malaspina, Sharif, & Amador, 2000) and older populations (Grady et al., 1995). There is also evidence showing that faster encoding usually results in faster and better retrieval (Kent & Lamberts, 2006; Lamberts, Brockdorff, & Heit, 2002). However, whether the ease of processing due to spatial congruency is robust enough to determine retrieval performance is not obvious.

The second prediction is that spatial incongruency rather than congruency will result in superior memory. This prediction might sound paradoxical; however, memory retrieval is fundamentally different than semantic judgement tasks reviewed above. Distinctiveness, in particular, has a substantial impact on verbal memory (see Hunt & Worthen, 2006 for a comprehensive review). Distinctiveness in the present study can occur and impact retrieval performance in two related ways:

Hirshman (1988) demonstrated that weakly related word pairs (e.g., quick - short) are better remembered than strongly related pairs (e.g., long - short). Weakly related word pairs are more likely to represent unexpected or novel semantic combinations and allow more elaborate

encoding than expected pairs. In turn, elaborate encodings give rise to more accurate retrieval.

In the current study, spatial words that are presented in incongruent locations can be assumed to violate participants’ expectations due to spatial simulations. Consequently, spatially incongruent words might enjoy a similar memory advantage over spatially congruent words due to reconstruction of more detailed encodings.

Craik and Lockhart (1972) suggested that the memory of an item is a function of the depth to which the item was initially processed. In keeping with this argument, Craik and Tulving (1975) reported higher retrieval accuracy for words processed at deeper levels which were induced with semantic questions at encoding (e.g., “Does ‘apple’ fit into the ‘fruits’ category?”) as to words encoded with questions related to phonological (intermediate level) or orthographic (shallow level) properties of the word. We expect that participants encode all words at semantic level regardless of spatial congruency (see Chapter 4). That said, spatially incongruent words could be encoded at a relatively deeper level due to possibly additional resources (e.g., attention, cognitive effort) required to process an unexpected situation. In this respect, unexpectedness can lead to deeper and more elaborate encodings, which, in turn, could lead to a memory advantage for words encoded in incongruent locations.

In sum, evidence reported above demonstrates that words with spatial associations can give rise to simulations of concerned locations. Simulated locations interact with where the words are perceived and in return, either facilitate or delay the processing of spatial words. Previous studies suggest a congruency advantage in memory performance. However, incongruency between physical and semantic space at encoding could lead to better memory performance because unexpected items are typically remembered more accurately.