Mismatch negativity (MMN) as a tool for investigating auditory discrimination and sensory memory in infants and children

Abstract

For decades behavioral methods, such as the head-turning or sucking paradigms, have been the primary methods to investigate auditory discrimination, learning and the function of sensory memory in infancy and early childhood. During recent years, however, a new method for investigating these issues in children has emerged. This method makes use of the mismatch negativity (MMN), the brain's automatic change-detection response, which has been used intensively in both basic and clinical studies in adults for twenty years. This review demonstrates that, unlike many other components of event-related potentials, the MMN is developmentally quite stable and can be obtained even from pre-term infants. Further, MMN amplitude is only slightly smaller in infants than is usually reported in school-age children and it does not seem to differ much from that obtained in adults. MMN latency has been reported to be slightly longer in infants than in adults but reaches adult values by the early school-age years. Child MMN does not seem to be analogous to adult MMN, however. For example, contrary to the results of adult studies, a prominent MMN can be obtained from in all waking- and sleep states in infants. Moreover, MMN scalp distribution seems to be broader and more central in children than in adults.

Introduction

During the 1990s, the mismatch negativity (MMN), an automatic change-detection response, has been one of the most intensively studied components of event-related potentials (ERPs) (for a review, see e.g. Näätänen, 1992, Kraus and Näätänen, 1995, Lang et al., 1995, Näätänen and Alho, 1997, Näätänen and Tiitinen, 1998). It is a response to any discriminable change in the stimulus stream (Näätänen et al., 1978, Näätänen et al., 1980, Näätänen and Michie, 1979, Näätänen et al., 1982). Although it has been suggested that MMN is specific to the auditory system (Näätänen, 1992), there is some evidence of a response resembling the MMN also in the visual (Cammann, 1990, Czigler and Csibra, 1990, Alho et al., 1992, Woods et al., 1992) and somatosensory systems (Kekoni et al., 1996).

The MMN-generating mechanisms react to a change between standard and deviant stimuli. Thus, MMN is not just a response generated by new, non-refractory afferent elements activated by an occasional infrequent stimulus. This statement is based on several pieces of evidence. First, MMN requires a stimulus change (e.g. at least two different stimuli) in order to be elicited. It is neither elicited by the first stimulus in a series (Cowan et al., 1993) nor is it obtained with a very long inter-stimulus-interval (ISI) or when the deviant stimuli are presented alone without intervening standard stimuli (Näätänen, 1985, Sams et al., 1985, Näätänen et al., 1987, Mäntysalo and Näätänen, 1987, Lounasmaa et al., 1989). Second, MMN can be elicited not only when stimulus intensity, duration, or ISI are increased, but also when they are reduced (Ford and Hillyard, 1981, Näätänen et al., 1989a, Näätänen et al., 1989b, Näätänen et al., 1993b). Thus MMN does not occur due to sensory adaptation to the standard stimuli while the deviant stimuli engage ‘fresh’ neural elements. Third, MMN can be elicited by the omission of an element of a compound stimulus or of the second of two paired stimuli if the ISI is short (Yabe et al., 1997). Fourth, MMN latency and duration are relatively long for minor stimulus changes, which is atypical of basic afferent responses (Näätänen et al., 1989a, Näätänen et al., 1989b).

Thus, MMN can be considered as an outcome of a ‘comparison process’ between a new, deviant stimulus and a memory trace formed by the standard stimulus in the auditory system (see e.g. Näätänen, 1990, Näätänen, 1992). Consequently, it has been proposed that the MMN reflects the operation of auditory sensory memory (Neisser, 1967). This memory storage enables one to code and maintain pre-attentively highly accurate modality-specific information but only for a relatively short period of time (Neisser, 1967, Näätänen, 1992, Cowan et al., 1993, Tiitinen et al., 1994).

By varying the ISI, one can determine the time span of auditory sensory memory, since a stimulus change cannot elicit an MMN if the memory trace of the repetitive stimulus has already decayed (Näätänen, 1992). In adults, the temporal dynamics of auditory sensory memory have been investigated in several MMN studies (Cowan et al., 1993, Mäntysalo and Näätänen, 1987, Näätänen et al., 1987; Winkler, and Näätänen, 1992), showing that MMN can be recorded with inter-stimulus-intervals (ISIs) ranging from a continuous tone (Lavikainen et al., 1995) to about 10 s (Böttcher-Gandor and Ullsperger, 1992, Sams et al., 1993). The MMN amplitude decreases as the interval between the stimuli is prolonged, demonstrating the ‘weakening’ of the memory trace as a function of time (Mäntysalo and Näätänen, 1987, Sams et al., 1993). When interpreting these results, it should be taken into account, however, that some recent studies claim that the maximal pre-deviant interval by which MMN can be elicited does not necessarily reflect the time-span of the underlying memory trace but the time by which the process underlying MMN still considers the preceeding standards as relevant context for the deviant (Näätänen and Winkler, in press). This interpretation is supported by the fact that no adult studies have reported a significant decrease in the MMN amplitude but rather they have reported the presence of MMN at a certain ISIs and none at a longer ISI (Mäntysalo and Näätänen, 1987, Sams et al., 1993). Thus there seem to be an inconsistency between MMN data and auditory sensory memory. A single stimulus can form a representation in auditory sensory memory but at least two or three preceding standards are needed for MMN to be elicited (Näätänen, 1992). This inconsistency has been explained by Cowan et al. (1993) by suggesting that a sensory-memory trace can be either in active or inactive state with regard to possible MMN elicitation.

The MMN not only can be recorded to any suprathreshold physical change in, for instance, frequency, intensity, or duration of a sound (for a review, see Näätänen, 1992), but it also can be obtained in response to more complex changes such as changes in the spatial location of a sound source (Paavilainen et al., 1989) or to the partial omission of a compound stimulus (Nordby et al., 1991). Recently it has been demonstrated that changes in speech stimuli also elicit MMN (see e.g. Aaltonen et al., 1987, Aaltonen et al., 1992, Aaltonen et al., 1994, Sams et al., 1990, Kraus et al., 1992, Kraus et al., 1993b, Kraus et al., 1995a, Kraus et al., 1995b, Kraus et al., 1996; in press; Sharma et al., 1993, Cheour-Luhtanen et al., 1995, Cheour-Luhtanen et al., 1996, Korpilahti, 1996, Leppänen and Lyytinen, 1997, Näätänen et al., 1997, Pihko et al., 1997, Pihko et al., 1999, Cheour et al., 1997a, Cheour et al., 1998a, Cheour et al., 1998b, Leppänen et al., 1999).

MMN is likely to be composed of at least two subcomponents, one of which, a sensory- specific subcomponent, is generated in the auditory cortex (see e.g. Hari et al., 1984, Giard et al., 1990; for a review, see Alho, 1995, Kropotov et al., 1995) and the other in the frontal cortex (see Näätänen and Michie, 1979, Giard et al., 1990). The major MMN generator is located in the left and right auditory cortices although it is not yet completely clear whether this generator is located in the primary or secondary auditory cortex (see, however, Csépe et al., 1987, Csépe et al., 1989, Javitt et al., 1992, Javitt et al., 1995, Karmos et al., 1993). It has been suggested that the frontal MMN subcomponent is associated with the involuntary switching of attention to deviant stimuli (Giard et al., 1990). So far, there have been only a few reports on intracranial MMN recordings in humans. Kropotov et al. (1995), however, have recorded MMN directly from the temporal cortex of two adult patients. Recently, these types of studies also have been conducted in children. Liasis et al. (1999) carried out intraoperative recordings of MMN from the cortical surface in children with epilepsy as a presurgical evaluation. Stimuli were either tones differing in duration (25 vs. 75 ms.) or frequency (1000 vs. 700 Hz), or they were speech stimuli (/da/ vs. /ba/). Standard and deviant stimuli ERPs were highly localized to areas surrounding the Sylvian fissure, revealing components to detection as well as discrimination. These potentials reflected activation of different neuronal populations and may contribute to the scalp-recorded N1 and MMN.

MMN also may have subcortical components. For example, Csépe et al. (1989) have recorded MMN-like negativity in cats over the medial geniculate body (MGB) of the thalamus and the dorsal hippocampus to changes in tones. More recently, Kraus et al., 1994a, Kraus et al., 1994b have recorded MMN from the auditory thalamus of guinea pigs. MMN was obtained only in non-primary auditory thalamus and it was absent in the primary subdivision.

It has been shown that the MMN generating systems depend on physical characteristics of stimuli. Topographic and spatiotemporal modeling studies demonstrate that the MMN scalp distribution is different to changes in frequency, intensity, and duration. Thus the MMNs to different changes may be generated by different neuronal populations (Alho et al., 1991, Tiitinen et al., 1992, Giard et al., 1994).

MMN has been used in several studies to investigate neurophysiological plasticity following learning. These studies demonstrate that training or experience in a certain language environment can change MMN amplitude, latency or duration. Importantly, the results of behavioral studies can be related to the results of MMN studies.

Kraus et al. (1995c) trained adult subjects to discriminate between two similar-sounding variants of the phonemes /da/ and /ga/. After 1 week of training, the MMN amplitude, area and duration had all increased. Training effects were maintained 1 month following the last training session. Näätänen et al. (1993) trained adults to discriminate complex spectro-temporal patterns. The authors divided one training session into early, middle and late sessions each of which was followed by discrimination task. According to the authors, after only one session, MMN amplitude had increased and peak latency decreased.

The existence of language-specific memory traces has been demonstrated in Estonian and Finnish adults (Näätänen et al., 1997). The standard stimulus /e/ is a phoneme prototype (Kuhl et al., 1992) in both Finnish and Estonian. Two deviant stimuli were used: one was a prototype in both languages and the other occurred only in Estonian. In Finnish adults, the MMN was larger to the Finish vowel and attenuated to the Estonian vowel. In Estonians this attenuation did not occur with the Estonian vowel.

Recently, Winkler et al. (1999) showed that MMN reflects neuronal plasticity in foreign language learning. Finnish-speaking Hungarians, non-Finnish speaking Hungarians, and Finnish subjects were investigated. The standard stimulus was /e/ which is a prototype in both languages and the deviant stimuli were /ae/, which is a prototype in Finnish but not in Hungarian, and /y/, which is a prototype in both languages. The results showed that the vowel /y/ elicited MMN of equal amplitude in all groups. In contrast, /ae/ elicited MMN only in Finnish and Finnish-speaking Hungarians but not in the non-Finnish-speaking Hungarians. Thus, the results demonstrate that learning a foreign language is associated with neural sensory changes which are reflected by the MMN amplitude.

These studies demonstrate the existence of much longer lasting memory traces that can be stored in auditory sensory memory. It appears that when a complex stimulus is heard often enough, a long-term memory trace of it is built. After these long-term memory traces have been developed, they can be activated automatically. Nevertheless, the process of building these traces seems to demand active processing of the stimuli, at least in adults. To recognize the complex stimulus as the same heard before this long-term memory trace of it has to be activated. These long-term memory traces seem to serve as recognition patterns, a match process, in the analysis of the auditory environment and may play an essential role in, for example, speech perception (Näätänen and Tiitinen, 1998).

After two decades of intensive basic MMN studies in healthy, young adults, new data have applied MMN to the study of impaired perception in clinical populations although, at least so far, MMN has not been used as a diagnostic or prognostic measure in clinical practice. Yet, it has been suggested, (Kraus et al., 1995a, Ponton and Don, 1995) that MMN may provide objective information about the cochlear implants function. It also has been demonstrated that it is possible to assess the effects of aging (Pekkonen et al., 1993) and to investigate deficits in auditory processing and memory in Alzheimer's (Pekkonen et al., 1994) and Parkinson's (Pekkonen et al., 1995a) diseases. MMN has been shown to be attenuated in patients with unilateral lesions of the dorsolateral frontal cortex (Alho et al., 1994), in schizophrenic adults (Shelley et al., 1991, Javitt et al., 1993, Javitt et al., 1998) as well as in patients with aphasia (Aaltonen et al., 1993). Importantly, Kane et al., 1993, Kane et al., 1996 has shown that, in comatous patients, the appearance of MMN was a positive prognostic sign that preceded waking up by 1–2 days. Patients who did not show MMN did not regain consciousness. Therefore, the MMN appears to be an early indicator of recovering from coma. Despite the large number of MMN studies in clinical populations, surprisingly few studies have concentrated on investigating the inter- and intra subject replicability and variability of MMN. Pekkonen et al. (1995b), however, have demonstrated in healthy adults that MMN amplitude and duration showed significant test-retest stability both at the group and individual level although MMN showed considerable interindividual variation. McGee et al. (1997) have also investigated variability of MMN in individual subjects. Nevertheless, considerable work needs to be done in this area before it can be utilized in mainstream clinical practice.