Lecture 8: Animal cognition - comparing models of working memory

cognitive abilities from a comparative perspective

-comparing human and animal abilities gives us a much clearer picture of our own abilities

cognition (range of definitions)

-basis for intelligent behaviour

-overrides reflexive, habitual response in favour of complex, long term goals

-controls sensory, memory and motor systems

-applying top down mental processes

features of cognition

-key structure is prefrontal cortex

-effort and attention required

-core vs higher-order cognitive abilities

core cognitive abilities (executive functions)

-foundational to higher order cognition

• working memory → hold and process information simultaneously 

• inhibitory control → withholding responses 

• cognitive flexibility 

higher order cognitive abilities

• object permanence → world exists around us and not just in our perspective 

• self-recognition → recognise ourselves and our position in the world 

• mental time travel → going back through our memories 

• theory of mind → our actions can affect others, understanding other’s perspectives 

• tool use/casual reasoning 

scala naturae (ladder of being) 

-hierarchal degree of perfection 

-at some point animals stop being simply instinctual and become introspective 

-based on how similar we feel things are to us (intelligence-wise)

order of scala naturae

-bottom-up from least intelligent to most intelligent:

1. fish

2. amphibians

3. reptiles

4. birds

5. mammals

-humans are at the top → anything observed as being close to us is seen as more intelligent

-evolutionary distant animals perceived as less intelligent

working memory in fish (evidence against scala naturae)

-showed fish stimuli that they had to learn and respond to

-fish demonstrated they had working memory

cognitive flexibility in frogs (evidence against scala naturae)

-frogs demonstrated cognitive flexibility and learned inhibition

-adapted to different situations

-could learn arbitrary rules and respond to them

inhibitory control in lizards (evidence against scala naturae)

-lizards can engage in inhibitory control tasks 

-demonstrates to a degree, a lack of instinctual behaviour 

limitation of scala naturae

-insight-related cognition might not be specifically human

cortex

-believed to be what differentiates humans from animals

-humans have a laminated cortex

laminated cortex

-6 layers of different types of cells that overlap

-lamination leads to rapid communication and is unique to humans

-defines a new computationally advantageous module

-differentiation of neocortex led to development of cognitive skills

prefrontal cortex (PFC)

-associated with core cognition and executive function 

-damage to regions associated with deficits in executive function 

-hard to dissociate between executive functions and where they are localised in the brain (use of lesions in patient studies)

size of PFC

-bigger in humans

-the bigger the PFC the more intelligent behaviours we see

-lamination in neocortex of PFC important for human behaviours

birds vs humans (comparative approach) 

-birds do not have a neocortex 

-have a pallium

-compare birds and humans as want to see if cortex is responsible for behaviour → so compare to animal without cortex  

working memory

-representation of items held in consciousness during experiences or after retrieval of memories

-short lasting and associated with active rehearsal or manipulation of information

-key for storage and controlled processes

-key structure is PFC

-have STM and WM

action potentials (mechanism for studying animal cognition)

-electrochemical signal

-brief, stereotypical changes in the membrane potential of a neuron

-due to opening and closing of ion channels

-due to flux of sodium and potassium

-governed by gated ion channels

action potentials in PFC (mechanism for studying animal cognition)

-measure neurons in the PFC or comparative areas

-see whether they’re responding to WM tasks

-measure action potentials in the form of spike trains

measuring action potentials (mechanism for studying animal cognition)

-inserting electrodes into brain during tasks

-measure voltage changes in extracellular space around neurons 

-gives indication of many different neurons ‘spiking’ 

-can visualise these spike trains with dots/dashes

-very precise timing of action potentials → spatial accuracy 

Niki - method (neural basis for WM)

-delayed response task for monkeys

-one of two cue lights is illuminated for 1s → left or right

-cue illumination is turned off for 2-3s (delay period)

-monkey has to choose between left and right to identify which light came on

-monkey has to hold onto information during delay period → trained monkeys predicted to perform better on this task

• if monkey uses PFC to hold onto information should see activity in PFC

Niki - results (neural basis for WM)

-initial spike at presentation of cue 

-sustained spike trains during delay → delay activity 

-delay activity → persistent firing rate change → bridges time gap of delay period 

-neurons sustaining the memory of which light was active before selection is made 

-robust finding → delay activity replicated in many studies 

dopamine and classical conditioning 

-when animals are conditioned to respond to certain stimulus 

-can see observable dopaminergic neurons firing 

-when given a reward at start of conditioning → neurons will fire more once they receive reward 

-once paired with conditioned stimulus → respond to conditioned stimulus instead of reward 

-so respond to anticipation rather than actual reward

neuromodulator (dopamine)

-links to delay activity (WM)

-responds to sensory stimuli which should predict future rewards

criticism of conditioning research

-neuronal activity might not be reflective of WM → may just be the brain’s response to anticipating a reward

-neuronal activity happens because been conditioned to respond in specific environment

not measuring WM (criticism of Niki)

-dopamine cells respond to sensory cues that predict reward 

-this signal could be used to tag sensory cues as relevant and facilitate their entrance into WM 

-so in the delayed response task: 

1. cue-left and cue-right lead to a response in dopamine neurons 

2. increases dopamine levels in prefrontal cortex

3. enables persistent delay activity in PFC 

monkeys not using WM (criticism of Niki)

-task does not involve any manipulation or interaction with information 

-only have to remember which light was on 

-so measures STM not WM → passive store of information 

delay activity

-activity is delayed

-so when processing information should see delay between stimulus and response

avian brain

-birds do not have cortex

-pallial layer → similar to human subcortical structures like basal ganglia

-primitive structures

-smaller areas considered to be akin to cortex in mammals

analogue

-same or similar function even if the structure is different

avian nomenclature forum

-reconsidered areas to be considered part of pallial area

-this area may support higher order cognition

-this is analogous to human cortex in terms of function → but fundamentally different to architecture of our cortex

pallium

-neuroanatomical term for the grey and white matter covering the cerebrum 

nidopallium caudolaterale (NCL)

• nidopallium → nested pallium 

-nidopallium has many subregions and one of them is the caudolateral part → the NCL 

-caudolateral means of the side and tail

investigating NCL (is NCL analogues to PFC?)

-WM task in homing pigeons

-directed forgetting procedure

-single unit recording in NCL during task performance

Rose - overview (is NCL analogues to PFC?)

-trained pigeons on a particular task

-put in front of a screen and presented with a single stimuli they had to learn

-done across two different trial types:

1. pigeons told to remember

2. pigeons told to forget

Rose - remember trial (is NCL analogues to PFC?)

-start with inter-trial interval (break)

-presented with a sample (circle stimuli)

-pigeon has to encode specific pattern/array of patterns

-pigeon receives a high note that instructs them to remember the sample information → will later be asked to recall it 

-3s delay period 

-engage in comparison → presented with an array and must discriminate between two targets and select what most closely matches what they had observed 

-if they are correct they receive a reward

Rose - forget trial (is NCL analogues to PFC?)

-start with inter-trial interval (break)

-presented with a sample (circle stimuli)

-pigeon has to encode specific pattern/array of patterns

-pigeon receives a low note that instructs them to forget the sample information → will not be asked to recall it 

Rose - results (remember trial)

-increase from sample

-consistent from cue to end of delay

-when provide stimulus see initial spike in neuron activity (encoding stimulus)

-when receive cue telling to remember information see sustained spike trains all the way through to the point that they need to make the choice

-means the pigeon is holding onto the information and maintaining it in the NCL → may be basis of WM in pigeons

Rose - results (forget trial)

-found that when the cue (telling pigeon to forget) is presented, the neural activity returns to baseline

-so the delay activity stops

-shows that when a pigeon is holding onto WM and then is instructed to let that information go, it is no longer being maintained

-NCL is locus of WM behaviour → so pigeons have WM even with completely different brain architecture to human PFC

Rose - abolished delay activity (issues with interpretation)

-abolished delay activity could be linked with reward prediction

-neurons could change firing rate in response to forget, because there is no reward

-difficult to discern between confounds of WM and reward prediction

-neurons could be involved in both

Rose - motor response (issues with interpretation)

-neuron activity could also be due to preparation of motor response

-forget cue requires no motor response

-so there are several possible interpretations

Veit - more advanced measures (follow-on study from Rose)

-used more advanced measures to disentangle WM from motor preparation

-presented with a sample and delay allows us to disentangle WM and motor preparation → every trial the crow has to make a choice, so motor preparation is happening in every trial

• therefore the study is only measuring WM

-matching to sample → have to hold onto the information until they have to make a choice

-randomised reward → neutralises the possibility of anticipation of the reward as it is unpredictable

Veit - method (working memory in crows)

-crow in space with a TV screen and electrode in NCL

-presented with ‘go stimulus’ → tells them they are about to engage in a task

-presented with pre-sample black screen

-then presented with a sample and then a delay → after delay have to make a choice about the sample they saw from a selection of options

Veit - results (working memory in crows)

-crows perform task at very high level

-good at cognitive tasks → perform close to 100 most of the time

Veit - sample-selective neurons (results)

-delay activity is not in response to the task itself

-crows demonstrate a preference in the firing rate of the neurons

-neurons fire differently in terms of their rates depending on what they are observing

-can see firing rate for each item across 4 trials → can see significant differences in the firing rates in response to what they are observing

-can see that crows hold onto relevant information but also respond differently to different stimuli that is presented to them to accurately complete the task

Veit - sample-selective neurons continued (results)

-preferences and differences in firing rate when holding and identifying information compared to during the delay before making their decision

-the activity is not arbitrary → specific to the stimuli the crows are observing

-sample selective neurons demonstrate preference during the delay period

-these differences in firing rate cannot simply be down to them performing the task → because there are significant differences depending on the type of stimuli they’re presented with

Veit - delay-selective neurons (results)

-the response of the neuron is selective during the delay period

-during the sample period the neuron did not distinguish the stimuli

Veit - population of neurons (results)

-more than half were either sample or delay selective or both

Veit - firing rate (results)

-analysis on relation between neural activity and task performance

-firing rate predicts whether or not they are likely to get the task correct

-in error trials the firing rate of the neurons does not distinguish the different stimuli as well as in correct trials

-true for sample and delay selective neurons

advantages of investigating cognition in animals

-animals studies allow direct recordings of action potentials → single cell recordings

-invasive and not usually possible in humans

-spatial and temporal accuracy can’t get from EEG or fMRI

-provide great insights into the nature of cognition, analogues of cognition in different species

limitations of investigating cognition in animals

-invasive - can be stressful for animals

-animals cannot self-report and have to be trained on tasks that do not reflect typical behaviour

-difficult to design tasks to remove confounds such as reward anticipation, motor preparation, STM or WM

bee brain

-0.5mm³

-only 1,000,000 neurons compared to 16,000,000,000 in humans

-not a simple, hard-wired structure

Zhang - delay period (WM in honey bees)

-to create delay period varied the length of the tunnel the bee had to fly through

-the closer the sample is to the entrance, the longer the bee has to fly (delay period) before it can make its choice

-vary the length of the tunnel and where the sample appears, can change how long a bee has to hold off something in memory

Zhang - method (WM in honey bees)

-bee flies through tunnel

-in the tunnel the bee is shown a sample (geometric shape)

-bee has to remember this sample

-where this sample is placed in the tunnel varies (increase/decrease delay period)

-as it progresses through the tunnel the bee has to make a choice between lots of different stimuli at the end of the tunnel (match-to-sample task)

Zhang - results (WM in honey bees)

-high performance levels for short delays (75% correct for 1.24s)

-performance indicates WM of up to 6.5s

Zhang - second method (WM in honey bees)

-added an incorrect pattern framing the other two patterns → so bee has to encode a pattern whilst also ignoring other patterns

-present with sample they have to ignore, then the correct sample and then another sample they have to ignore → correct sample is always shown in the same position, so bees have to learn this and ignore the other stimuli

-called a transfer task

Zhang - second results (WM in honey bees)

-perform pretty well on this task

-only have an issue when shift the pattern → bees don’t know what they are looking for so they fail

-but if it is kept consistent they will start to perform well again

limitations of Zhang (WM in honey bees)

-showed the bees can maintain and withdraw information to get the correct answer

-this is argued to reflective of WM

-however ignoring stimuli may not be manipulating memory → may just be flexibility or another cognitive process