Intonation and positional effects in spoken serial recall

Intonation and Positional Effects in Spoken Serial Recall Michelina Savino (michelina.savino@uniba.it) Dipartimento di Scienze della Formazione, Psicologia, Comunicazione, Universita di Bari, piazza Umberto I, 1 70121 Bari, ITALY Andrea Bosco (andrea.bosco@uniba.it) Dipartimento di Scienze della Formazione, Psicologia, Comunicazione, Universita di Bari, piazza Umberto I, 1 70121 Bari, ITALY Martine Grice (martine.grice@uni-koeln.de) IfL-Phonetik, Universitat zu Koln, Herbert-Lewin-Strasse 6 50931Cologne, GERMANY Abstract Past studies have indicated that intonation, in the sense of fundamental frequency modulation, can only enhance serial recall to the extent that it can induce a grouping effect, something that can also be induced by a simple insertion of pauses. However, in a study of spoken serial recall of nine- digit lists, we are able to show that recall is significantly better when sequences of digits are marked by specific intonation contours than when they are simply grouped by silent pauses in the signal. Thus, we found that intonation plays a role during the encoding phase, whereby items in group-final positions draw particular benefit from intonation. However, intonation does not appear to play the same role in the retrieval phase, since when subjects are instructed to imitate intonation during recall, performance shows mixed effects. Keywords: serial recall; intonation; grouping effect; short- term memory Introduction In field research, there is a general consensus that serial recall (short-term memory in general) and prosody are closely related. Well-documented evidence of such a relationship is the grouping effect, that is, the enhanced recall of items in a list when they are presented in groups (for example, Reeves et al 2000). The grouping effect is stronger for auditory stimuli (Cowan et al 2002, Frankish 1985), as prosody plays an important role in this grouping, or patterning. Past and more recent research has aimed to ascertain the nature of these groups. In their seminal work on the auditory grouping effect, Frankish (1995) and Saito (1998) provided evidence that it can be obtained by temporal organisation of speech stimuli realised by pause insertion between groups as well as by superimposing a “natural” intonational pattern, or by manipulating pitch levels on groups of items (Frankish 1995). On the other hand, more recent studies have shown that the grouping effect in serial recall reflects rhythmic groups, referred to as the stress grouping effect, rather than intonational phrases (Reeves et al 2000, Boucher 2006, Gilbert & Boucher 2007). It is argued that these groups correspond to the segmentation units (chunks) listeners use in spoken language perception (Gilbert et al 2011). Previous research dealing specifically with the role of intonation in improving serial recall suggests that it is relevant to the extent that it can induce a similar grouping effect to that obtained by pause insertion (Frankish 1995, Saito 1998). However, a number of potentially relevant aspects of intonation deserve further exploration. For instance, in the previous studies discussed above, the superimposition of a fundamental frequency (F0) contour 1 on the whole sequence, or F0 manipulation on groups of items within a sequence were carried out with little control over specific tunes and their associated meanings. This is particularly relevant, since the role of intonation in signalling discourse structure is widely acknowledged, as it cues hierarchical relationships among phrases within a discourse unit (Hirschberg & Pierrehumbert 1986). Moreover, in sequences intonation can convey information about the hierarchical structure (groups) as well as about specific positions within a group. In Italian (in particular the variety of Bari), a rich inventory of tunes is available for marking those kinds of hierarchical relationships in a sequence (Savino 2001; 2004), among which the most typical are: - The “continuation rise” contour, a gradually rising F0 movement from the nuclear syllable up to the end of the phrase. It signals that the list has not been completed yet, and that more items are to come (“non-finality” contour); - A high rising contour, where the rise in F0 starts before the nuclear syllable and continues rising up to the end of the phrase. It conveys the information that the current item is the penultimate in a sequence, i.e. that the end of the list is approaching (“pre-finality” contour”); - A falling contour, involving a gradual fall from the nuclear syllable until the end of the phrase. This contour marks the end of a sequence (“finality” contour). Our aim here is to verify whether the use of specific tunes conveying such hierarchical relationships and positional The F0 contour corresponds roughly to what is perceived as the pitch, or melody.