Implicit learning of what comes when and where within a sequence: The time-course of acquiring serial position-item and item-item associations to represent serial order

Abstract

Much research has been conducted aimed at the representations and mechanisms that enable learning of sequential structures. A central debate concerns the question whether item-item associations (i.e., in the sequence A-B-C-D, B comes after A) or associations of item and serial list position (i.e., B is the second item in the list) are used to represent serial order. Previously, we showed that in a variant of the implicit serial reaction time task, the sequence representation contains associations between serial position and item information (Schuck, Gaschler, Keisler, & Frensch, 2011). Here, we applied models and research methods from working memory research to implicit serial learning to replicate and extend our findings. The experiment involved three sessions of sequence learning. Results support the view that participants acquire knowledge about order structure (item-item associations) and about ordinal structure (serial position-item associations). Analyses suggest that only the simultaneous use of the two types of knowledge acquisition can explain learning-related performance increases. Additionally, our results indicate that serial list position information plays a role very early in learning and that inter-item associations increasingly control behavior in later stages.

Keywords: SRT; chaining; implicit sequence learning; race model; serial order.

Figure 1.

Structure of one trial (Panel A), one mini block (Panel B), and one session (Panel C). A: An example of one trial is shown. In each trial, participants had to search for a tilted T among rotated Ls and press a button that corresponded to the tilt of the T (left or right). Please note that the target is encircled only for purposes of illustration; during the experiment there was no circle around the target. B: After each fourth trial, a fixation cross appeared on the screen and stayed on for 1,000 ms. The regular response-stimulus-interval (RSI) was 400 ms. C: In each session, five learning blocks were followed by one transfer block. In each learning block, fixed sequence and random sequence mini blocks appeared in random order. For details on the structure of the transfer blocks see text, see Figure 2 and Table 1.

Figure 2.

Schematic illustration of memory structures of fixed sequences (Panel A) and derived transfer sequences (Panels B-D). In all cases, encircled letters correspond to elements of a sequence, with sequential presentation going from left to right. The boxed numbers above the sequence elements indicate representations of the respective serial positions. Arrows correspond to associations. A: In our view, sequence learning results in the formation of item-item as well as of position-item associations. The former are indicated by the round arrows between sequence elements, the latter by the straight arrows between the serial positions and the sequence elements. In the learning blocks, two repeated fixed sequences could be learned. It is important to note that participants learned two different sequences, A₁-B₁-C₁-D₁ and A₂-B₂-C₂-D₂. The italic letters indicate a sequence element and the indices the sequence identity. Therefore, A₂ corresponds to a different target screen location than A₁, etc. B: To test for position-item associations, the ordinal-only sequences feature trials that have not been used during learning (indicated as n), as well as test trials where a target screen location from one of the learned sequences occupied the same serial position, n-n-C₁-n. (Element C₁, now being the third element in the sequence, as in the upper part for Panel A.) C: Only item-item association information is available. In this case, an order-only trial needs to be preceded by the same sequence element as it is during the learning phase. For example, in the sequence C₁-D₁-C₂-A₁, element D₁ is preceded by element C₁ as during the learning phase (importantly, C₁ and D1 both are from the same, but C₂ is from a different sequence, as mirrored by the indices), so the reaction time (RT) during the trial with element D₁ is considered (see Panel C). D: Situations where no associative knowledge could be used for prediction/retrieval facilitation. In this case A₂ is now preceded by D₂, unlike in the learning phase. Hence the RTs in the trial where the target appeared at screen location A₂ are considered. Please note that unlike in the examples, the test item appeared at all possible serial positions, not only at the third serial position. Analyzing trials where two target screen locations appear in the learned order at the wrong serial position can provide insights into item-item associations.

Figure 3.

Development of reaction times (RTs) during the learning phase. The figure shows mean RTs from the fixed sequence (solid circles) and random (empty circles) conditions as a function of block. Vertical dashed lines indicate the beginning/end of a session (about 48 hr without training). Bars indicate standard errors for within-subject designs (based on the interaction effect, see Loftus & Masson, 1994).

Figure 4.

Reaction times (RTs) in different transfer sequence conditions relative to RTs in fixed sequences and new-location trials. A: For each transfer block, the RTRT difference between test trials in new-transfer sequences (ordinal-only) and fixed sequences (gray bars) and new-location trials (black bars) are shown. Panels B and C also show the two respective differences, where in Panel B order-only trials are taken as reference and in Panel C control trials. For further descriptions, see text.

Figure 5.

Estimated cumulative density functions for order, ordinal, and fixed sequence trials. The figure shows the 10 estimated percentile points for each of the four functions of interest: G_order, G_ordinal, G_{fixed sequence}. The figure also displays the calculated G estimates of G_order + G_ordinal, which is central to assess the validity of the race model inequality. RT = reaction time.